DISCUSSION PAPER SERIES NO. 2018-26
DECEMBER 2018
Poverty is Multidimensional: But Do We Really Need a Multidimensional Poverty Index?
The PIDS Discussion Paper Series constitutes studies that are preliminary and subject to further revisions. They are being circulated in a limited number of copies only for purposes of soliciting comments and suggestions for further refinements. The studies under the Series are unedited and unreviewed. The views and opinions expressed are those of the author(s) and do not necessarily reflect those of the Institute. Not for quotation without permission from the author(s) and the Institute.
CONTACT US:RESEARCH INFORMATION DEPARTMENTPhilippine Institute for Development Studies
18th Floor, Three Cyberpod Centris - North Tower EDSA corner Quezon Avenue, Quezon City, Philippines
[email protected](+632) 372-1291/(+632) 372-1292 https://www.pids.gov.ph
Jose Ramon G. Albert and Jana Flor V. Vizmanos
Poverty is Multidimensional:
But Do We Really Need a Multidimensional Poverty Index?
Jose Ramon G. Albert Jana Flor V. Vizmanos
PHILIPPINE INSTITUTE FOR DEVELOPMENT STUDIES
December 2018
1
Abstract
The Philippines, just like many other developing countries, has measured poverty using one-
dimensional monetary-based indicators, although poverty is multidimensional in nature.
Various work on generating a multidimensional poverty index (MPI) has been undertaken
using data from several countries, including that done by the United Nations Development
Programme with the Oxford Poverty and Human Development Initiative. Several studies have
also used national data for generating an MPI. The measurement of multidimensional poverty
in the Philippines should depend on a careful investigation on whether there is much value
added in producing a composite index of poverty. Just because poverty is multi-dimensional
need not mean that its measurement should be. This study examines discusses various issues
regarding the production of a MPI, from the choice of the underlying indicators for the index,
the weights assigned to the indicators, as well as the aggregation process. It also reviews
measurements on quality of life (happiness and well-being), on poverty and welfare (including
multidimensional poverty) and on sustainable development. It examines various possibilities
of an MPI for the Philippines based on several waves of household surveys (viz., the National
Demographic and Health Survey, the Annual Poverty Indicator Survey, and the Family Income
and Expenditure Survey) and several approaches on choices of indicators and weights. This
paper also looks at the robustness of trends in the resulting MPI approaches, and some policy
issues attendant to the measurement of multidimensional poverty, especially on its relationship
with traditional income poverty measurement.
Keywords: multidimensional poverty; composite index; indicators; weights; aggregation
2
Table of Contents
1. Introduction ........................................................................................................ 4 2. Measurements Beyond GDP ............................................................................. 5
2.1. Measures of Happiness and Well-being ..................................................................... 6 2.2. Measures of Development and Progress .................................................................... 8 2.3. Traditional Poverty Measurement ............................................................................. 11 2.4. Measurement of Sustainable Development .............................................................. 14 2.5. Measuring Multidimensional Poverty ........................................................................ 15
3. Empirical approach for measuring multidimensional poverty ..................... 18 3.1. Choice of Indicators and Dimensions ........................................................................ 19 3.2. Choice of Weights .................................................................................................... 22 3.3. Identification of the Poor and Aggregation of Poverty Data ....................................... 23
4. Empirical Results ............................................................................................. 25 4.1. Trends ...................................................................................................................... 27 4.2. Robustness .............................................................................................................. 30
5. Summary and Policy Implications .................................................................. 34 6. References ....................................................................................................... 36
List of Boxes
Box 1. Goals 1 to 6 of the Sustainable Development Goals ............................................ 15
List of Tables
Table 1 Dimensions and Indicators of Deprivation Used in this Study ............................ 20
Table 2 Multidimensional Poverty Measures from the Global MPI Approach* ................. 25
Table 3 Share of Population by Poverty and Vulnerability Status .................................... 26
Table 4 Incidence of Deprivation (in %) from the Global MPI Approach* : 2008-2017 ..... 26
Table 6 Contribution to MPI by Dimension ...................................................................... 29
Table 7 Distribution of Filipinos by (Per capita) Income Cluster and by MPI-Poverty or
Vulnerability Status: 2015 .................................................................................. 30
Table 8 Distribution of Filipino households by (Per capita) Income Quintile and by MPI-
Poverty or Vulnerability Status: 2017 ................................................................. 30
Table 9 Multidimensional Poverty Measures Using Nested Equal Weights (NEW) and
Principal Components Analysis (PCA)-Based Weights : 2008-2017 .................. 32
List of Annexes
Annex Table A-1 Dimensions and Indicators of the Multidimensional Poverty Index. ...... 41
Annex Table A-2 Intensity of Deprivation, Multidimensional Poverty Headcount and
Proportion of Population Deripvied in Living Standards, Education and
Health Dimensions : 2009-2017. ......................................................... 42
Annex Table A-3 Income Poverty and Multidimensional Poverty Profiles by Various
Subpopulations: 2015. ........................................................................ 44
3
List of Figures
Figure 1 Trends in Real Per Capita based indicators of Income and Consumption (from
the National Accounts and Household Surveys) and in Poverty Rates (based
on International and National Poverty Lines) .................................................... 6
Figure 2 Structure of the Better Life Initiative Index. ......................................................... 7
Figure 3 Structure of the Human Development Index and Sub-Indices ............................. 9
Figure 4 Structure of the Social Progress Index. ............................................................. 10
Figure 5 Multidimensional Poverty Index versus Human Development Index, Social
Progress Index, GDP per capita and Poverty Rate (i.e., $1.25 per day) of
64 countries, various (recent) years. ................................................................ 17
Figure 6 Multidimensional Poverty Headcount and Monetary Poverty Headcounts......... 27
4
Poverty is multidimensional: But do we really need a multidimensional poverty index?
Jose Ramon G. Albert and Jana Flor V. Vizmanos*
1. Introduction
Economic growth is an important aspect of socio-economic development. Traditionally, the the
health of an economy is measured as the percent rate of increase in real Gross Domestic Product
(GDP), which, in turn, represents the value of a country’s aggregate output (goods or services
produced). When GDP is divided by total population, the resulting measure, called GDP per
capita, represents the potential income of each person in the population if the aggregate income
is equally shared. Neither GDP nor GDP per capita, however, provides a sense of how
resources and wealth are allocated across a society. Despite such limitations, the usefulness of
GDP as a measure of economic performance cannot be discounted as socio-economic
development is intertwined with economic performance. Economic growth enhances a
country’s potential for reducing poverty and solving other social and environmental problems.
The notion of development, especially sustainable development, is, however, much wider than
that of economic growth (CGD, 2008; Soubbotina, 2004). Development comprises both the
need and the means by which to provide better lives for people; development entails both
economic growth as well as progress in overall quality of life — say, in terms of health,
nutrition, education. Sustainable development is development that successfully balances
economic goals with social and environmental ones. While some developing countries over the
past half century have achieved high economic growth rates, narrowing the gap significantly
between themselves and the prosperous countries, but many more developing countries are not
catching up. Further, across the pages of history, we can find various examples of countries
where economic growth was not necessarily followed by progress in development of the quality
of life for the vast majority, where growth was achieved but at a cost of either greater inequality,
higher unemployment, overconsumption of natural resources, loss of cultural identity, or a
combination. Thus, we also need other measures to describe quality of life, progress and
sustainable development, other than GDP or GDP per capita.
Many developing countries, including the Philippines, have been measuring and monitoring
welfare (and poverty) in their respective societies, based on single money-metric terms, either
from consumption or income data (UNSD, 2005). In recent years, the National Economic and
Development Authority (NEDA) as well as the Philippine Statistics Authority (PSA) have
made public pronouncements1 that government is making steps to adopt a multidimensional
measure of poverty, owing to the recognition of poverty as having dimensions beyond income
poverty. Furthermore, consistent with the Filipino aspirations highlighted in AmBisyon Natin
* The authors are senior research fellow and research assistant, respectively, of the Philippine Institute for Development Studies (PIDS). The views expressed here are the authors’ own.
1 http://www.neda.gov.ph/2015/11/03/ph-reiterates-a-multidimensional-perspective-in-poverty-reduction/ ;
http://www.neda.gov.ph/2018/10/04/neda-wants-better-measurement-of-poverty/ ; https://www.unescap.org/sites/default/files/Session6.2.1_Philippines_Role_in_Developing_Indicator_Framework_and_SDG_Monitoring.pdf ; https://businessmirror.com.ph/psa-digs-deep-into-child-poverty-incidence-data/
5
2040 (NEDA 2015), the NEDA is also working toward the development of a Quality of Life
Index (QLI) through the collection of data in a pilot survey2.
This study aims to examine how much value added is there in producing such a measure of
multidimensional poverty, and what would this entail. In the next section, we firstly review
some measures of development beyond GDP, as well as describe the current measurement of
welfare and poverty in the country. In the third section, we get into the mechanics of
constructing a composite index for describing the multidimensional aspects of poverty. In the
fourth section, we provide and discuss empirical findings of various approaches to generating
a multidimensional poverty index (MPI) using waves of several household surveys of the PSA,
viz., the FIES, the National Demographic and Health Survey (NDHS), and the Annual Poverty
Indicator Survey (APIS). We end in the last section with a summary of learning lessons from
this study.
2. Measurements Beyond GDP
In 1990, the United Nations Development Programme (UNDP) released its first Human
Development Report (UNDP, 1990). From then up to 2015, the world has made amazing
advances in income growth (yielding an average annual GDP growth of 3.5 percent). Further,
the world has reduced the proportion of persons with incomes less than $1.25 a day in 2005
purchasing power parity prices from 47 percent in 1990 to 14 percent in 2015; it has also
improved the health, education, and living conditions of people across the world (UN, 2015a).
Despite these gains, progress across and within countries has been uneven.
Economic growth, while important for poverty reduction, is not sufficient, as growth is not
equivalent to sustainable development, to improvements in well-being, and to inclusive
opportunities for social mobility. For instance, the Philippines has undergone a different
economic growth trajectory in the past decade, but this economic growth has not yet translated
into substantial income poverty reduction (Figure 1). We certainly need measures of progress
in a broad sense other than indicators, such as the GDP, Gross National Income, or even the
unemployment rate.
2 https://www.philstar.com/headlines/2018/10/08/1858200/neda-poverty-index-chart-pinoy-standard-living
6
Figure 1 Trends in Real Per Capita based indicators of Income and Consumption
(from the National Accounts and Household Surveys) and in Poverty Rates (based on
International and National Poverty Lines)
Source: Philippine Statistics Authority (PSA)
2.1. Measures of Happiness and Well-being
Recognizing the limitations of the Gross National Income (i.e., which is GDP plus net primary
income from abroad) as a measure of development, Bhutan’s Former King Jigme Signye
Wangchuck first conceived of the idea of measuring Gross National Happiness (GNH) in 1972.
Since 2008, Bhutan has thus far conducted three GNH surveys covering all twenty districts
(Dzonkhag) of the country (CBS and GNH Research 2015). The first GNH survey
questionnaire collected data about living conditions and religious behavior of respondents.
Later rounds of the GNH Survey used a shortened instrument, but the surveys retained
questions on religious behavioral. The latest of the GNH Surveys was conducted in 2015,
which suggested that 91.2% of the Bhutanese were happy, and that the index increased to 0.756
in 2015 from 0.743 in 2010. The measurement of GNH revolves around examining four pillars
of happiness, viz.,
(1) promotion of equitable and sustainable socioeconomic development;
(2) preservation and promotion of cultural values;
(3) conservation of natural resources; and
(4) establishment of good governance.
across thirty-three indicators on quality of life from nine domains:
(1) community vitality;
(2) Cultural diversity and resilience;
(3) Education;
(4) Health;
(5) Psychological well-being;
(6) Time Use;
(7) Living Standards;
(8) Ecological diversity and resilience;
7
(9) good governance.
The indicators comprising the GNH Index thus include socio-economic indicators on living
standards, health and education as well as other aspects of quality of life, such as culture and
psychological wellbeing (CBS and GNH Research 2015; Ura et al. 2012). Following ideas
espoused in Alkire and Foster (2011), reports on the GHN identify four groups of people –
unhappy, narrowly happy, extensively happy, and deeply happy, they explore the happiness
people enjoy already, as well as suggest policies that can increase happiness and sufficiency
among the unhappy and narrowly happy people. The GNH Index is disaggregated by
meaningful sub-populations, such as men and women, as well as by district.
The idea to measure happiness has gained attention and interest among governments (both
national and cities) and various organizations, even in the private sector. The OECD, for
instance, has been publishing biennial reports since 2011 that discuss its examination of well-
being and societal progress using the Better Life Initiative Index (OECD 2011). The
conceptual framework of the index draws on the report of the Stiglitz-Sen-Fitoussi Commission
(2009); it distinguishes between current and future well-being (Figure 2). Current well-being
is described with two broad domains: material living conditions (income and wealth, jobs and
earnings, housing conditions); and quality of life (health status, work-life balance, education
and skills, social connections, civic engagement and governance, environmental quality,
personal security and life satisfaction). Following the approach recommended by the UNECE-
Eurostat-OECD Task Force on Sustainable Development (UN, OECD, Eurostat 2008), future
well-being (or sustainability of well-being) is measured through indicators of different types of
‘capital’ that drive well-being over time (OECD 2015).
Figure 2 Structure of the Better Life Initiative Index.
Source: (OECD, 2015)
8
In the United Kingdom, the Office for National Statistics (ONS) developed a framework for
measuring national well-being in 2010 consisting of 40 headline indicators from 10 domains.
The domains identified by ONS were the natural environment, personal well-being, our
relationships, health, what we do, where we live, personal finance, the economy, education and
skills, and governance. The ONS conducted a national debate between November 2010 and
April 2011 to find out ‘what matters’ to individuals and also to engage with experts on well-
being who would provide insight into what to measure and how to measure it. The
ONS collected over 30,000 responses during the debate which was conducted by holding
events across the UK, an online debate and engaging with the public via a variety of social
media.
In the Philippines, the Social Weather Stations (SWS) has been conducting surveys that ask
respondents information on happiness and life satisfaction since 1991. Its latest Fourth Quarter
2017 SWS Survey that tracks happiness puts happiness of Filipinos at a record high of 94%
reporting to be “very happy/fairly happy” with life in general (SWS 2018). Further, the SWS
has been collecting data on happiness together with 44 other countries to generate the
Happiness Index in the World Happiness Reports (SDSN 2018).
While there is growing interest in the idea to measure happiness even among official
statisticians in the Philippines (Virola et al. 2010), most national statistics offices across the
world however have not adopted the idea of measuring happiness and understandably so, since
the framework for the system of national accounts (on which GDP and Gross National Income
are based) itself has taken decades for countries to develop, while a concept like happiness has
nuances across cultures. Unlike other measures of happiness and quality of life, the GNHI in
Bhutan, for instance, involves religious behavior measurement components. Undoubtedly, the
relative importance of the different dimensions of happiness and quality of life will vary if
these were to be adapted in other countries. Further, the selection of indicators used to monitor
achievements in the dimensions of these composite indices measuring happiness and quality
of life may also differ when adopted across countries as there will be country specificities that
consider culture, history and measurement challenges. For issues on the ultimate usefulness of
measures of happiness or life satisfaction, see Graham (2009) and Wolfers (2008).
2.2. Measures of Development and Progress
In its Human Development Reports (HDRs) that have been released since 1990, the UNDP
discusses the Human Development Index (HDI)3, a summary measure of average achievement
in key dimensions of human development, i.e., health, education and standard of living. The
framework for HDI focuses on people, their opportunities and choices. The advantage of using
the HDI is that it describes in a single measure how countries have performed in attaining
overall human development. The disadvantage of the index is that, as any aggregate composite
index (such as those that attempt to measure happiness and well-being), it does not allow us to
see the relative importance of the different components of the index, or to understand why the
value of the index changes over time.
3 http://hdr.undp.org/en/content/human-development-index-hdi
9
Figure 3 Structure of the Human Development Index and Sub-Indices
The health dimension of the HDI is assessed by life expectancy at birth, while the education
dimension is measured by the mean of years of schooling for adults aged 25 years and expected
years of schooling for children of school entering age. Finally, the standard of living dimension
of HDI is measured by gross national income per capita. The scores for the three HDI
dimension indices are then aggregated into a composite index by way of a geometric mean.
The HDI enables comparison of countries with similar level of development but with different
human development outcomes.
During the 2000 Conference of the International Association of Official Statistics on
“Statistics, Development and Human Rights” held in Montreux, Switzerland, many official
statisticians and experts across the world expressed concern about the usefulness of the HDI
(OECD 2001). While there was recognition that the HDI is quite useful for advocacy, but it
was criticized as not being as useful for policy since policy priorities will still have to be
determined sectorally (Ravallion 2010). That is, as was earlier pointed out, when the HDI
changes, an examination of the components of the HDI will still be necessary to determine
what has changed and what has not. In addition, it has not been easy to justify how to put
weights to the components of a composite index.
Another yardstick, the Social Progress Index (SPI)4, has been developed by the Social Progress
Imperative under the technical guidance of Professors Michael Porter from Harvard Business
School and Scott Stern from the Massachusetts Institute of Technology. The SPI measures a
comprehensive array of components of social and environmental performance and aggregates
them into an overall framework. Similar with the HDI, the SPI is based upon social outcomes,
which determines the level of social progress achieved within a particular country. The stark
difference is on the inclusion of other indicators such as institutional, environmental, equity
and inclusion factors, among others.
4 http://www.socialprogressimperative.org/data/spi
10
Figure 4. Structure of the Social Progress Index.
The SPI is based on three dimensions, each with four components. These three dimensions
include basic human needs (such as nutrition and basic medical care, water and sanitation,
personal safety and shelter); foundations of well-being (indicated by access to basic knowledge,
access to basic information and communication, health and wellness, and ecosystem
sustainability); and opportunity (echoed by personal rights, access to higher education, personal
freedom and choice, equity and inclusion). Each of the components of the three dimensions
have a certain number of indicators that describe the components. All in all, fifty-four
indicators are currently used to form the SPI.
However, there is also lot to be desired in the selection of the fifty-four indicators of the SPI.
As in the case of other composite indicators, what justifies the selection and use of the fifty-
four indicators in the SPI?
Equal weights are given to the indicators for each of the twelve components because as the SPI
report says “there is no clear theoretical or empirical reason to weight any of the components
more highly.” For instance, the Access to Information and Communications component of the
index has four indicators that include fixed broadband subscriptions, internet users, mobile
telephone subscriptions, press freedom index. The Health and Wellness component considers
six indicators which includes life expectancy, obesity, cancer death rate, deaths from HIV,
deaths from cardiovascular disease and diabetes, and availability of health care. It is a puzzle
why fixed broadband would be effectively given ¼ weight, but yet, life expectancy, would be
giving a 1/6 weight. Why would cancer deaths be given the same weight as life expectancy,
and deaths from HIV?
Social Progress Index
Basic Human Needs
Nutrition and Basic Medical Care
Air, Water and Sanitation
Shelter
Personal Safety
Does a country provide for it's peoples most essential needs?
Foundations of Well Being
Access to Basic Knowledge
Access to Information and Communications
Health and Wellness
Ecosystem Sustainability
Are the building blocks in place for individuals and
communities to enhance and sustain well being?
Opportunity
Personal Rights
Access to Higher Education
Personal Freedom and
Choice
Equity and Inclusion
Is there opportunity for all individuals to reach their full
potential?
11
2.3. Traditional Poverty Measurement
To develop proper policy instruments for reducing poverty, a country must have a credible
poverty measurement system. Three essential steps comprise traditional poverty measurement
and diagnostics: (a) identifying an indicator of the welfare of households (and consequently
all members of the household); (b) setting a poverty line, a minimum acceptable standard of
that welfare indicator; and (c) aggregating the poverty data (Haughton and Khandker 2009;
Albert 2008; UNSD 2005).
Welfare Indicator. Developing countries that measure poverty commonly use are monetary
measures of welfare, either based on household income or household consumption. In the
Philippines, the welfare indicator used in the official poverty measurement system is per capita
income, sourced from the triennial Family Income and Expenditure Survey (FIES), conducted
by the Philippine Statistics Authority (PSA).
While many developing countries use consumption/expenditure as their welfare indicator for
poverty measurement (UNSD 2005), the Philippines uses income, as do China and Malaysia.
The use of income data for poverty metrics has its strengths given there are fewer number of
sources of income than the number of items for consumption/expenditure, thus, it is
operationally easier to collect total income of a household. But using income also has
limitations since income data is likely to be underreported due to memory recall biases, the
reluctance of respondents to reveal accurate information due to tax purposes or because some
income may be from illegal sources (Haughton and Khandker 2009). Furthermore, the
accuracy of certain components of total income, such as agricultural income, cannot be assured
as this would depend on when data collection was undertaken (i.e., whether before or after the
harvest). The extent of biases in income measurement is, however, likely to be high on the
upper tail of the income distribution, whose effect is not of particular concern in poverty
measurement and analysis.
Analysts generally view consumption-based measures of poverty as providing a more adequate
picture of well-being than those based on income, especially in low- or middle-income
countries (Haughton and Khandker 2009; UNSD 2005). Typically, income fluctuates across
months, and even from year to year. It also rises and falls in the course of one’s lifetime whereas
consumption remains relatively stable (and is thus viewed to be a better measure of permanent
income than income itself). Further, consumption may be more accurately measured than
income as survey respondents may be more able and willing to recall what they spent rather
than what they earned, especially if more detailed questions jog or push the respondent’s
memory. The extent and direction of biases of reported expenditure is however unclear: the
possibility of prestige bias on those in the lower-part of the expenditure distribution cannot be
discounted.
There are also issues that complicate the aggregation of total expenditures, especially on how
to account for consumption on durable goods, as well as how to measure the value of home
production and home services.
Jogging memory from the use of detailed questionnaires may also have its limitation:
respondents may suffer from information fatigue after hours of being asked detailed questions
on their expenditures. The entire FIES module takes an average of five hours of interview per
household, with the household visited twice—in July, to obtain the first semester information,
12
and in January of the following year to get the second semester information on family income
and household expenditures (Albert 2008).
In most cases, we expect consumption poor households to also be income poor (and vice versa),
but some consumption-poor households may have high income, and some income-poor
households may have high consumption. Thus, it is far from clear whether income-based
measures of poverty are less superior to consumption/expenditure-based measures of poverty.
What is only clear is that there is no perfect indicator of well-being, and that each monetary
measure of poverty has its strengths and limitations.
Poverty Lines. Poverty lines should represent what is required to purchase a bundle of essential
goods (typically food and nonfood items) to maintain a minimal standard of well-being. While
there have been attempts to adopting a standard methodology across countries in setting
national poverty lines (UNSD 2005), but there has been no full consensus because of the belief
that ultimately, national poverty lines are somewhat arbitrary and need to resonate with social
norms. Typically, the food (component of the) poverty line is set with the cost of basic needs
method, which entails determining the price of some nutritional benchmark through an artifice.
In most countries, the artifice is a basket of generic food items, benchmarked to daily energy
requirements of around 2100 kilocalories of energy per person (Albert and Molano 2009).
The differences in methodologies in the choice of a welfare indicator, the approach for data
capture, and the setting of poverty lines across countries make cross-country poverty
comparisons with national poverty lines contentious.
To monitor global poverty, the World Bank currently uses $1.90 in purchasing power parity
poverty (PPP) 2011 prices. This poverty line essentially means converting the equivalent of
one US dollar and 90 cents to a local currency based on 2011 PPP exchange rates and updating
this by inflation. The PPP exchange rates essentially capture the cost of living difference among
countries. But criticisms have been raised against this approach. For example, Reddy and
Pogge (2008) point out that the use of the international poverty lines is not adequately anchored
on the real cost requirements of purchasing basic necessities.
Aggregating Poverty Data. One of the typical aggregates of poverty data is poverty incidence,
i.e., the proportion in poverty, which may be derived for both households or the entire
population. The poverty incidence is a simple measure for assessing overall progress in
reducing poverty. A weakness though of this poverty rate is that the depth or intensity of
poverty experienced by poor people and poor households are not taken into account. Other
poverty measures such as the poverty gap and poverty squared gap can be produced for such
purposes. However, these indices, especially the poverty squared gap, are not easy to interpret;
hence, they are hardly used for practical field work.
Official Poverty Statistics in the Philippines. According to Republic Act 8425 of 1997 (Social
Reform & Poverty Alleviation Act), those who are “poor” are “individuals and families whose
income fall below the poverty threshold as defined by the NEDA and/or cannot afford in a
sustained manner to provide their minimum basic needs of food, health, education, housing
and other essential amenities of life.” Thus, this definition recognizes many dimensions of
poverty, such as health, food and nutrition, water and environmental sanitation, income
security, shelter and decent housing. The PSA (specifically, one of its predecessor statistical
agencies, the National Statistical Coordination Board) has been releasing official poverty
statistics based on the triennial FIES since 1985.
13
In the Philippines, the official food poverty line is estimated at urban and rural areas of each
province by using a one-day food menu as an artifice for setting official poverty lines. These
menus satisfy energy, and other nutrient requirements. The official poverty methodology
consists of constructing the menus first with a national menu, rather than the previous approach
of having varying menus across the regions, with provincial prices to satisfy a daily food
requirement (Virola 2011). In addition, a constant Engle’s coefficient is used in the current
methodology for indirectly estimating the non-food component of the total poverty line across
urban/rural areas in each province. This makes the estimation consistent across the country,
compared to the previous methodology.
In 2012, official poverty statistics based on first semester income data sourced from the FIES
were released and compared to the corresponding statistics for the first semesters of 2006 and
2009. A year later, poverty data were also generated sourced from the APIS, which is conducted
by the PSA on non-FIES years. Prior to 2013, the APIS collected income and expenditure data,
but using a less detailed questionnaire than the FIES. Although the 2013 APIS used more
questions on income (than it used to) with its 19 pages of questions, the 2012 FIES income
module used 24 pages of questions. However, even if APIS 2013 made use of the entire 24-
page income module of FIES 2012, this would still not make the resulting income data from
the APIS and FIES comparable since FIES also asks households detailed information on their
expenditures before income questions are asked, using a questionnaire with a s length of 78
pages (that takes an average interview time of 5 hours to accomplish). The NEDA and PSA
have compared the 2013 APIS-based poverty data, but trends cannot actually be obtained from
the APIS and the FIES given the lack of full comparability of the survey instruments (Albert
et al. 2015). At best, comparisons can be made within waves of a household survey, i.e., APIS
with APIS, or FIES with FIES.
While traditional poverty statistics have been simple headline summaries of poverty conditions,
they have their limitations. It is not enough to use poverty rates across areas (such as countries
and regions within a country) for resource allocation, since total population varies across areas.
In the Philippines, some areas such as the Autonomous Region of Muslim Mindanao (ARMM)
may have very high poverty rates but the number of poor persons in ARMM is actually much
smaller than in some regions where poverty incidence figures are lower but where the total
population is much higher. Further, even as poverty rates for a population can be generated by
assuming that all members in a poor household are poor, the disaggregation of poverty statistics
by sub-groups, e.g., males and females, may not necessarily capture the actual differences in
gender disparities given that intra-household differences are often not captured in traditional
poverty measurement.
As has been pointed out earlier, poverty is a multidimensional phenomenon. Poor people view
their poverty much more broadly than income or consumption poverty, to include lack of
education, decent employment, health, housing, empowerment, personal security. In the next-
subsections, we discuss the global indicators for monitoring sustainable development and the
MPI. Some studies, e.g., Gwatkin et al. (2000); Filmer and Pritchett (2001) have also looked
into developing a deprivation index, a weighted composite index of poverty indicators (largely
asset data), by way of principal components analysis, and have used such an index instead to
monitor (asset-based) poverty.
14
2.4. Measurement of Sustainable Development
Over the years, there has been recognition that not all development paths are sustainable.
Various definitions of sustainable development have been developed (See, e.g. Pezzy, 1992 for
a review). Behind these concepts and definitions is the recognition that economic development
can erode human and natural capital. To be sustainable, development must provide for all assets
(physical, human and natural capital) to grow over time—or at least not to decrease.
Thus, the World Bank has been examining “development diamonds” to examine the
relationships among life expectancy at birth, gross primary (or secondary) enrollment, access
to safe water, and Gross National Income per capita for a given country relative to the averages
for that country’s income group, i.e., low-income, lower-middle income, upper-middle-
income, or high-income group (Soubbotina, 2004). Each of the four socio-economic indicators
is put on an axis, then connected with bold lines to form a polygon. The shape of the resulting
development diamond is then compared to a reference diamond, which represents the average
indicators for the country’s income group, each indexed to 100 percent. Thus, any point outside
the reference diamond shows a value better than the group average, while any point inside
signals below-average performance.
Further, the World Bank, as well as the United Nations5, have been encouraging countries to
account for changes in a country’s natural capital (i.e. valuation of the environment) in
calculations of the national accounts (particularly indicators such as GDP and the Gross
National Income) in order to explore sustainable development issues. The Wealth Accounting
and the Valuation of Ecosystem Services (WAVES) 6 partnership led by the World Bank aims
to promote sustainable development by ensuring that natural resources are mainstreamed into
development planning and national economic accounts. Several indicators such as genuine
domestic savings rate and genuine domestic investment rate are also being monitored by the
World Bank. These indicators adjust the traditional domestic saving rate and genuine domestic
savings rate downward by an estimate of natural resource depletion and pollution damages (the
loss of natural capital), and upward by growth in the value of human capital (which comes
primarily from investing in education and basic health services). Recently, the World Bank
has also come up with a human capital index that combines indicators of health and education
into a measure of the human capital that a child born today can expect to obtain by her/his 18th
birthday, given the risks of poor education and health that prevail in the country where s/he
lives (Kraay 2018).
In September 2015, 194 countries, including the Philippines, committed to attaining the 17
Sustainable Development Goals (SDGs) and their 169 targets by 2030 (UN, 2015b). The SDGs
aim to work on the unfinished agenda of the Millennium Development Goals (MDGs) that
were launched in 2000, with a more ambitious set of targets. Over a year after the SDGs were
launched, chief statisticians across the world agreed on an indicator framework of 232
5 A major step towards accounting for natural capital in the national accounts came with the adoption by the UN Statistical
Commission of the System for Environmental and Economic Accounts (SEEA) in 2012. This provides an internationally‐agreed method to account for material natural resources like minerals, timber and fisheries. For more information, see https://seea.un.org/ 6 For information on the Wealth Accounting and the Valuation of Ecosystem Services Project of the World Bank, see http://www.worldbank.org/en/news/feature/2015/06/15/waves-faq
15
indicators7 for monitoring the extent of meeting the SDGs, including the eradication of extreme
poverty, but without resorting to using a composite index on sustainable development.
Further, the SDGs, particularly the first six global goals covering poverty reduction, as well as
quality education for all, health and nutrition, gender equality, safe drinking water and safe
sanitation:
Box 1. Goals 1 to 6 of the Sustainable Development Goals
SDG1 End poverty in all its forms everywhere
SDG2 End hunger, achieve food security and improved nutrition, and
promote sustainable agriculture
SDG3 Ensure healthy lives and promote well-being for all at all ages
SDG4 Ensure inclusive and equitable quality education and promote
life-long learning opportunities for all
SDG5 Achieve gender equality and empower all women and girls
SDG6 Ensure availability and sustainable management of water and
sanitation for all
suggest that poverty has many “forms” beyond mere monetary deprivation. The recognition
of poverty as being multidimensional is rooted in viewing poverty as “capability failure” (Sen
1999). With poverty viewed as multidimensional, we can look into a range of specific
indicators of capabilities including those relating to health, education, shelter, and access to
basic amenities to capture the multiple deprivations of poor people. The key issue is whether
income (or consumption) offers an adequate representation of this range of capabilities, and if
it did, then there would not really be much value added for a separate measurement on
multidimensional poverty. However, just because poverty is multi-dimensional need not mean
that its measurement should be. The 232 global SDG indicators, for instance, or even subset of
the available indicators forms a dashboard not only on sustainable development but also on
multidimensional poverty.
2.5. Measuring Multidimensional Poverty
In 2010, drawing from methodological work done at the Oxford Poverty and Human
Development Initiative (OPHI), the HDR introduced the MPI, an overall headline indicator of
poverty that enables poverty levels to be compared across places and over time in order to see
at a glance which groups are poorest and whether poverty has been reduced or has increased
(UNDP 2010; Alkire and Foster 2011).
Subsequent HDRs since 2011 have released the UNDP estimates of multidimensional poverty,
with adjustments documented in their methodological reports. In 2014, an innovative MPI
(MPI-I) was also developed in the HDR to explore improvements in the original approach
(MPI-O) to estimate MPI (Kovacevic and Calderon 2014). The 2014 and 2015 HDRs contained
7 In March 2016, the United Nations Statistical Commission (UNSC) approved a list of 230 indicators for monitoring the SDGs.
A year later, the UNSC revised the list to 232 indicators (https://unstats.un.org/sdgs/indicators/indicators-list/ ). See also the 2017 IAEG-SDGs report to the UNSC (https://unstats.un.org/unsd/statcom/48th-session/documents/2017-2-IAEG-SDGs-E.pdf )
16
both MPI-O and MPI-I estimates. The MPI-O was aligned with indicators used to track the
Millennium Development Goals, the predecessor global agenda of the SDGs.
The UNDP and OPHI have recently developed a new version of the global MPI this 2018 with
five of the ten indicators revised to align the MPI with the SDGs; the new estimates on global
MPI take account of data coverage, communicability, comparability, disaggregation, and
robustness (Alkire et al. 2018). The global MPI for 2018 covers 105 countries, which comprise
nearly four-fifth (77 %) of the world’s population (corresponding to 5.7 billion people); of this
proportion, a quarter (23 %) of people (amounting to 1.3 billion) are identified as
multidimensionally poor. In contrast, the World Bank’s current estimate of the poor (earning
less than the international poverty line of $1.90 in 2011 PPP prices) globally is about a tenth
(10.1%). For the Philippines, the estimate of MPI poor based on the global MPI methodology
is 7.4%, while the 2015 estimate of (consumption) poverty rate using the international poverty
line is at 8.3%.
Multidimensional poverty measurement follows the same steps as traditional poverty
measurement: choosing indicators to represent dimensions of deprivation; setting thresholds
(or cutoffs) with these indicators and dimensions, and, aggregating the poverty data to
summarize information on individual deprivations for the population. In order to further guide
debates and designs of development policy, the MPI identifies deprivations across the same
three dimensions of the HDI on health, education and living standards. Unlike HDI, the MPI
is based on 10 indicators, two representing health (malnutrition, and child mortality), two are
educational achievements (years of schooling and school enrolment), while six are indicators
of “living standards” (including access to electricity, sanitation, safe drinking water, and
proxies for household wealth, such as type of floor, cooking fuel, and some asset ownership).
Each dimension is weighted equally, and within a dimension, each indicator is given equal
weights. Annex Table A-1 lists the ten indicators for the global MPI and their actual
descriptions (including changes in the indicators across the years).
While the HDI uses aggregate country-level data, the MPI makes use of micro data from
household surveys, which are then aggregated to a national measure of multidimensional
poverty. That is, the MPI assesses poverty at the individual level. If someone is deprived in a
third or more of the ten (weighted) indicators, then s/he is ‘multidimensionally poor’, and the
intensity of her/his poverty is measured by the number of deprivations s/he is experiencing.
Following the methodology of Alkire and Foster (2011), the MPI is calculated by multiplying
the incidence of poverty (H) and the average intensity of poverty (A), with the latter averaged
only among those considered poor. That is
MPI = H × A,
Thus, the MPI reflects both the share (or incidence) of people in poverty as well as the degree
to which the poor people are deprived. The usefulness of the MPI methodology is that aside
from generating the incidence of multidimensional poverty, it allows an examination of the
prevalence (i.e., how many people experience overlapping deprivations) as well as the intensity
(i.e., how many deprivations people experience at the same time) of deprivation, and for
specific dimensions. Unlike the conventional monetary-based measures of poverty, the MPI
enables policymakers to have information on various dimensions of poverty by showing
interconnections among the various aspects where poor are actually deprived.
17
Datt (2017) pointed out that: “multidimensional poverty comparisons are sensitive to
assumptions in relation to the choice of indicators, the weights assigned to the indicators, the
dimensional (deprivation) and the overall poverty cut-offs, as well as the choice of the
aggregate poverty measure.” Another major issue regarding the MPI is that the indicators were
chosen not necessarily because they are the best available data on each of the three broad
dimension of poverty, but because the MPI methodology requires that a poverty analyst has all
the indicators for exactly the same sampled person or sample household. That is, the indicators
must all come from a single survey. Further, is there any extra value added policy use for the
MPI given that when the MPI changes, we still need to examine the specific component
dimensions and indicators, or are such measures going to be more useful for advocacy?
Figure 5 shows that the MPI, is very strongly correlated with the SPI, HDI, and even with
monetary poverty rates, although there are a few outliers for the latter. This is somewhat
expected since these composite indices involve indicators that are correlated with consumption
or income, but these indicators, though, are not very likely to change much when there are
economic fluctuations, such as global economic slowdowns or upswings in macro-economic
performance. These indicators smooth out fluctuations in the MPI.
Figure 5 Multidimensional Poverty Index versus Human Development Index, Social
Progress Index, GDP per capita and Poverty Rate (i.e., $1.25 per day) of 64 countries,
various (recent) years.
Sources: Social Progress Imperative (http://www.socialprogressimperative.org/data/spi), United Nations
Development Programme (http://hdr.undp.org/en/data), World Bank (http://data.worldbank.org)
Datt (2017) and Balisacan (2015) have also pointed out that recent trends in income poverty
in the Philippines have been puzzling in the wake of fairly robust economic growth starting
2012, suggesting that there may be weaknesses in the current official poverty measurement
system in the country. Related to this, Albert et al. (2017) suggested three reasons for the
18
seeming puzzle : (a) the incidence of growth has not been pro-poor (i.e., high levels of income
inequalities have made economic growth largely benefit the high income classes, thus
minimizing the effects of growth on reducing poverty); (b) the updating of official poverty
lines (at the provincial urban/rural levels) by the PSA has overstated the cost of living in the
country; (c) there has been divergence in national accounts-based and survey-based growth in
per capita income and expenditure. The second reason is not a major explanation because trends
in official poverty that the PSA releases do not differ from overall trends in World Bank’s
estimates of (consumption) poverty that involve international poverty lines of USD 1.9 per
person per day in 2011 PPP prices (see Figure 1). The first and third reasons are also not
mutually exclusive. Thus, while the puzzle about high GDP growth and the lack of income
poverty reduction may actually be explained, and cannot be used to justify the need for a
multidimensional measure of poverty in the Philippines. Birdsall (2011) suggests that here are
three intrinsic reasons for multidimensional measure of poverty: “technical policy rationale (to
contribute to more effective policies at the technical level); the conversation-changer rationale
(to alter the discourse on what matters in the first place); and the advocacy rationale (to
communicate better, whether to acquire new or stronger advocates for change, or to name and
shame relevant actors).” That poverty is multidimensional coupled with the need to explore the
interconnected links of the many dimensions of poverty, and the need to have better actions to
yield better development outcomes are the central arguments for exploring an MPI, as is to be
undertaken in the next sections.
3. Empirical approach for measuring multidimensional poverty
The previous section provided a review of various composite indicators of welfare, happiness,
and progress. In this section we discuss the data and methodology used in this paper for the
possible measurement of multidimensional poverty in the Philippines bearing in mind broad
issues about construction of composite indices, viz., choice of indicators, weights and
aggregation (Ravallion 2012; Ravallion 2011, Alkire et al. 2015; Birdsall 2011; Ferreira and
Lugo 2013; and Bourguignon and Chakravarty 2003). We note that there is hardly any
disagreement among poverty analysts that poverty is multidimensional, that traditional poverty
measurement is imperfect, and that the multiple domains of deprivation are conceptually and
often correlated. What experts seem to disagree on is how best to measure poverty: just because
poverty is multidimensional need not mean we should measure it multidimensionally with a
single index. There are other ways of communicating the multidimensional nature of poverty
beyond a composite index such as through cross-tabulation dashboards and visualizations of
these dashboards. The parsimony of composite indices, whether the MPI, HDI or measures of
happiness, is appealing to some extent —reducing multiple dimensions into a single aggregate,
but the meaning, interpretation and robustness of these indices needs probing for these to be
useful and convey value added especially as each dimension/indicator component has
measurement errors.
As was earlier mentioned, this study makes use of waves of three household surveys conducted
by the PSA, viz., (a) the NDHS; (b) the FIES; (c) the APIS. It examines more closely the
robustness of results across the different data sets used in the next section. Together with the
Demographic and Health Survey or Multiple Indicator Cluster Survey of countries, the NDHS
has been used as the data set for the global MPI (Alkire et al. 2018). The usefulness of this
survey is that it has a wealth of health (and mortality) information, aside from education and
asset data (of households and household members). Alternative MPI specifications to the
global MPI value for the Philippines have been developed by Datt (2017), Bautista (2017) and
Balisacan (2015) for the Philippines using either the APIS or the FIES. While the APIS and
19
FIES do not have anthropometric and mortality information, but these surveys have income
and expenditure data and can thus be used to link monetary poverty data with nonmonetary
dimensions of poverty (Ericta and Luis 2009; Ericta and Fabian 2009). As was pointed out
earlier, income data in the APIS in recent years has become more detailed, leading the PSA to
yield income poverty statistics from the APIS, though these statistics are incomparable to those
sourced from the FIES (Albert et al. 2015). The APIS and FIES are also
The triennial FIES, the APIS and the quarterly Labor Force Survey (LFS) follow an integrated
survey programme through a master sample design. Sample households across household
surveys and survey rounds follow a rotation scheme, to minimize respondent fatigue. For the
quarterly LFS, one rotation of the sample households is dropped every quarter and replaced by
a new set of sample households from the respective sample areas. The FIES and APIS are
riders to the LFS. For the quarters when the FIES is a rider to the LFS, a semester later, the
same households are visited to get the second semester information for the FIES and also to
conduct the LFS. Since the FIES and APIS are riders to the LFS, some of the household
member information from the LFS (such as educational attainment and employment) may also
be merged with the FIES and APIS to yield deprivation indicators (although employment is
not used in the NDHS-based indicators for the global MPI).
The NDHS, FIES and APIS were designed to generate reliable estimates of indicators up to the
regional level. Since these surveys are conducted for different purposes and vary in the
deprivation indicators, even for the same variable of interest (e.g., food expenditure in APIS
and food expenditure in FIES), comparisons of deprivation indicators and resulting MPIs have
to be taken with a grain of salt.
3.1. Choice of Indicators and Dimensions
The choice of dimensions and indicators for the construction of any composite index is guided
by a conceptual framework and data availability. Several implementations of multidimensional
poverty measurement for the Philippines (e.g., Datt 2017, Bautista 2017, Balisacan 2015),
including the global MPI (Alkire et al. 2018) make use of the three dimensions of poverty
pertaining to education, health and standard of living (Annex Table A-1). In this paper, we
continue making use of these three dimensions, partly to see the extent of consistency with
results from these previous work, and partly to examine the robustness of trends if different
indicators were to be used.
For the global MPI, the final list of 10 indicators covering the three dimensions (Annex Table
A-1) were selected after a consultation process involving experts in all the three dimensions,
an examination of data availability and of cross-country comparison issues (Alkire et al. 2018).
In this study, all indicators (shown in Table 1) used for constructing multidimensional poverty
measures reflect socio-economic welfare. The choice of indicators, however, had to depend on
indicator availability from the household survey being used. With multidimensional poverty
aimed at expressing the joint distribution of deprivations across different dimensions, a key
data consideration is the ability to examine deprivations across three dimensions of education,
health and living standards for the same set of households or individuals. On one hand, this
might seem to be a limitation, as there would be no way to combine information from other
surveys. On the other hand, this can also be considered a strength as empirical results allow us
to see interconnections among the component dimensions and indicators. Since the indicators
20
of deprivation varied in availability in the NDHS, FIES, APIS, the estimates of
multidimensional poverty were expected to vary.
Table 1 Dimensions and Indicators of Deprivation Used in this Study
Dimension Deprivation
indicator
Indicator criteria : household is
considered deprived if
NDHS FIES* APIS*
education school
attendance
any child aged 5-17 is not attending
school
education years of
schooling
no member had educational
attainment of elementary graduate or
better
health child mortality any child aged 0-5 died
health food
consumption
food expenditure is less than food
poverty threshold
living
standards
electricity no electricity
living
standards
sanitation toilet facility is not water-sealed,
sewer septic tank/other depository,
closed pit and/or shared with other
households
living
standards
source of water water source is not from community
water system (own or shared),
tubed/piped deep well (own or shared)
or protected spring
living
standards
cooking fuel household cooks with dung, wood or
charcoal
living
standards
housing
materials (roof
and walls)
housing materials for roof and walls
are not strong
living
standards
tenure status household resides in a housing unit/lot
with no consent of the owner
living
standards
assets household does not own
a) a durable (e.g. television123,
radio123, washing machine23,
refrigerator23, stove/oven/
microwave oven23, aircon23,
personal computer23) or
communications asset (e.g.
landline123, mobile phone123)
and
b) a mobility asset (e.g.,
car/truck123,
motorcycle/tricycle/bicycle123)
Notes: *= merged with data from Labor Force Survey (LFS); 1 = available in NDHS; 2 = available in FIES; 3 = available in APIS
21
While the global MPI makes use of 10 indicators, only 8 are available in the NDHS. All of
these eight NDHS indicators except the floor materials indicator were used in this study,
together with two other welfare indicators, viz., housing materials (which is available in all
three surveys) , and tenure status (which is also found in FIES and APIS).
The selection of the 10 deprivation indicators for the global MPI was guided mainly by expert
discussions on common practices, especially in the context of the MDGs and SDGs. The latter
consideration suggests that the set of deprivation indicators varies across the three household
surveys. For example, there are more deprivation indicators linked with standard of living in
both FIES and APIS than in NDHS. Furthermore, APIS also collects information about the
experience of hunger (but the manner of questioning was not the usual practice in CSOs that
collect hunger data for the 2014 APIS). Coverage of households members for health insurance
is also asked in APIS and NDHS, but the manner of asking in early years for the APIS was not
for all household members. Due to the question wording issues, we opted not to consider using
hunger and health insurance indicators for this study.
For the education dimension, two deprivation indicators are used in this study: (i) the years of
schooling of household members (which is available across the three surveys) and (ii) current
school attendance of school-age (i.e. aged 7-16 years) children (which is available in FIES and
APIS through the LFS) . A household is considered deprived of education functionings for the
first indicator if not one member of the household has completed basic education. For the
second indicator, a household is deprived of educational functionings if it has a school-age
child who is currently not attending school.
For health, four deprivation indicators used in this study are on child mortality, food
expenditure, hunger and health insurance coverage. Child mortality is only available in
NDHS, but it is proxied in APIS and FIES by other living standards indicators, viz., the lack
of access to safely managed water supply and sanitation services (which is also available in
NDHS). The experience of hunger is only available in APIS, but it is also proxied by food
expenditure, especially if this expenditure is less than the food poverty threshold. The lack of
health insurance by a household (available in NDHS and APIS) does not provide a pathway
for the household to manage risks to welfare from illnesses.
For the living standards dimension, eight indicators are used. Two mentioned earlier, viz,
access to clean water and to safe sanitation, proxy deprivation indicators on health. The
remaining indicators measure access to electricity, quality shelter (floor and materials for roof
and walls), clean energy for cooking, and assets (both mobility and non-labor assets, viz.
durables or communication assets). The indicators on floor and on clean energy for cooking
are available only in the NDHS. For this study, the deprivation indicators used was chosen to
be parsimonious and fairly comparable over time (although across waves, some changes may
have been made in survey instruments).
Datt (2017) also made use of indicators on employment, a dimension that is not in the global
MPI. There is sufficient justification for this given the effect of employment on income and
consumption. We also look into this separately to further examine robustness of estimates in
multidimensional poverty measurement. We however look go beyond his use of indicators
regarding unemployment, but also make use of indicators based on underemployment.
22
3.2. Choice of Weights
As regards the weights used to aggregate across indicators and dimensions for
multidimensional poverty measurement, Decanq and Lugo (2013) provide a review of various
approaches. Ravallion (2010; 2011; 2012) critiques the lack of an intrinsic meaning of the
associated weights in the MPI (and even the HDI) as regards prices, which are used to add the
components of consumption expenditure (or, incomes used to finance consumption) 8. Current
implementations of MPI generally adopt equal weights or a natural variant, viz., the nested
equal weights approach, where each dimension is given equal weight, then all indicators within
the dimension are also given equal share of the dimension weight. These approaches implicitly
assumes specific tradeoffs between the constituent components of welfare. The use of equal
weights and variants, or even the use of ad hoc weights is unable to explain ordering of
households according to multidimensional welfare, nor is it readily apparent how this is done
with the use of such weights. An extra amount of one component can offset the change in
another component and leave the index unchanged, but such tradeoffs are hardly stated,
explained and communicated explicitly.
A statistical approach for the assignment of weights involves the use of principal components
analysis (PCA)9. Several studies such as Gwatkin et al. 2000; Filmer and Pritchett 2001 use
8 Under the law of one price, and given relatively weak assumptions on consumer preferences, the relative prices are equal to the rate at which consumers— regardless of their income levels and allowing for different utility functions—are willing to trade one such component of the index (e.g., safe drinking water) for another (e.g., an asset such as television) 9 PCA is a multivariate statistical method that is primarily used to reduce a large set of correlated variables into a smaller set of uncorrelated variables while retaining as much of the variation in the original dataset as possible. From an initial set of n correlated and standardized variables, X1 through to Xn, PCA creates m uncorrelated indices or components, where each of the m new variables or variates is called a principal component (PC). Each PC is a linear weighted combination of the initial variables:
𝑃𝐶1 = ∑ 𝑎1𝑗𝑋𝑗
𝑛
𝑗=1
𝑃𝐶2 = ∑ 𝑎2𝑗𝑋𝑗
𝑛
𝑗=1
⋮
𝑃𝐶𝑛 = ∑ 𝑎𝑛𝑗𝑋𝑗
𝑛
𝑗=1
Standardized variables mean that the variables have a mean of zero, and unit variance; if the variables are unstandardized, they can be readily transformed into standard units by subtracting the mean and dividing the result by the standard deviation of the variable. PCA amounts to rotating the original standardized variable space to a point where the variance of the new variate (PC) is maximized.
The first PC is that unit length linear combination) of the initial variables X1 through to Xn that has the maximum variance among all unit length linear combinations of X1 through to Xn.
The second PC is that unit length linear combination. of the original variables X1 through to Xn that is uncorrelated with the first PC and has maximum variance among all among all unit length linear combinations of X1 through to Xn that are uncorrelated with the first PC.
The third PC is that unit length linear combination of the initial variables X1 through to Xn that is uncorrelated with the first two PCs and has maximum variance among all among all unit length linear combinations of X1 through to Xn that are uncorrelated with the first two PCs.
…
23
PCA to (standardardized units of) welfare indicators for deriving a “deprivation index” from
the first principal component. However, we merely make use of the re-scaled factor loadings
of the first principal component on the pooled sample from a particular survey as the alternative
weights for the indicators to generate the multidimensional poverty measures.
3.3. Identification of the Poor and Aggregation of Poverty Data
Aside from the choice of indicators and the selection of weights for the indicators, another
important issue in measuring multidimensional poverty is the identification and aggregation
process. Given the various indicators, how should the poor be identified, and how can
deprivations across households (or individuals) and dimensions be put together into a single
measure of multidimensional poverty?
As pointed out in Datt (2017), the identification of the multidimensional poor may be done
two ways: (a) the use of the cross-dimensional cut offs specified in terms of the minimum
percentage of (weighted) dimensions a person (or household) must be deprived in for the
individual (or household) to be considered poor (see Alkire and Foster 2011; UNDP 2010;
Alkire et al. 2018); (b) the union approach where a person (or household) is considered
multidimensionally-poor if deprived in any dimension (Balisacan 2015).
Both approaches assume that each of the m dimensions of poverty characterize the state of
well-being of n individuals (or households). An individual (or household) i, where 𝑖=1,…,𝑛, is
viewed to be deprived in dimension j, where 𝑗=1,…,𝑚, if the person (or household) falls below
some predetermined threshold 𝑧𝑗 for that dimension. That is, let 𝑥𝑖𝑗 represent the individual (or
household) i’s actual achievement in dimension j, then this person (or household) is considered
deprived in dimension j if
𝑥𝑖𝑗 < 𝑧𝑗
Let 𝐼𝑖𝑗be a binary (0-1) variable that denotes whether or not individual (or household) i is
deprived in dimension j. That is,
𝐼𝑖𝑗 = {1 𝑖𝑓 𝑥𝑖𝑗 < 𝑧𝑗
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Further, let 𝑤𝑗 be weights for the jth dimension of poverty, where 0 < 𝑤𝑗 < 1 and ∑ 𝑤𝑗𝑚𝑗=1 .
The overall deprivation score for each individual (or household) can be calculated as the sum
of the weighted deprivation scores
𝑐𝑖 = ∑ 𝑤𝑗𝐼𝑖𝑗𝑚𝑗=1
With the cross-dimensional cut-off approach, we can calculate censored deprivation scores of
all the n individuals (or households) can be calculated using this identification function:
𝑐𝑖(𝑘) = 𝜌𝑖(𝑘) 𝑐𝑖
The last PC is that unit length linear combination. of the original variables X1 through to Xn that is
uncorrelated with the first n - 1 PCs and has maximum variance among all among all unit length linear combinations of X1 through to Xn that are uncorrelated with the first m - 1 PCs.
24
where 𝜌𝑖(𝑘) is a binary (0-1) variable denoting whether (or not) individual or household i is
deprived in at least k-fraction of the weighted dimensions, i.e.,
The multidimensional poverty index (MPI) is defined as the average of the censored
deprivation scores of the total population
𝑀(𝑘) = 1
𝑛∑ 𝑐𝑖(𝑘)𝑛
𝑖=1 =1
𝑛∑ 𝜌𝑖(𝑘) 𝑐𝑖
𝑛𝑖=1 =
1
𝑛∑ ∑ 𝑤𝑗𝐼𝑖𝑗
𝑚𝑗=1 𝜌𝑖(𝑘)𝑛
𝑖=1 .
The MPI can be also conveniently rewritten as
𝑀(𝑘) = 𝑞
𝑛[∑
1
𝑞 𝑐𝑖(𝑘)
𝑛
𝑖=1
]
where q is the total number of poor people, i.e.,
𝑞 = ∑ 𝜌𝑖(𝑘)
𝑛
𝑖=1
Thus, the MPI can be viewed as the product of H (the headcount ratio ) and A (the intensity A
of poverty) where the latter is the average deprivation score of poor people. Because of this
decomposition of MPI, the index is also considered an adjusted headcount ratio, where A serves
as an adjustment that accounts for the breadth of poverty.
While Alkire and Foster (2011) allow the cross-dimensional cut-off to range from the minimum
weight of any dimension to 100 percent, the global MPI (UNDP 2010; Alkire et al. 2018) sets
the cut-off at one-third.
On the other hand, for the union approach, the multidimensional poverty measure is written
much more simply as
𝑀(𝑈𝑛𝑖𝑜𝑛) = 1
𝑛∑ ∑ 𝑤𝑗𝐼𝑖𝑗
𝑚𝑗=1
𝑛𝑖=1
where the poor are identified by reference to a cross-dimensional cut-off specified in terms of
the minimum percentage of (weighted) dimensions a person must be deprived in for him/her
to be considered poor
The difference between the multidimensional poverty incidence measures for the cross-
dimensional cut-off and the union approach is that while the union approach counts all
deprivations of all individuals, the cross-dimensional cut-off approach counts the deprivations
of only those who are deprived in at least k-fraction of all weighted dimensions. The union
approach asserts the essentiality of all deprivations.
Further, when transfers are made from a more to a less deprived person, the poverty measure
increases for the union approach. In this paper, we make use of multidimensional poverty
measures from both a cross-dimensional cut-off of one-third as well as union-based approach:
25
4. Empirical Results
Estimates of the MPI, as well as multidimensional poverty headcount (H), and average
deprivation intensity experienced by the poor (A), and other multidimensional poverty
measures using the (old) approach for the global MPI estimation are given in Table 2, together
with the average annual rate of change of these statistics for the period covered by the 2017
NDHS, 2013 NDHS and 2008 NDHS data.
Table 2 Multidimensional Poverty Measures from the Global MPI Approach*
Measures of
Multidimensional Poverty
from the Global MPI
Year Annual rate of change, %
2017 2013 2008
2017-
2013
2013-
2008
2017-
2008
Multidimensional Poverty
Index
(MPI = H*A)
0.021
0.033
0.035 -9.74 -2.25 -5.09
Headcount ratio (H):
Population in
multidimensional poverty (%)
4.3
6.3
6.8 -8.82 -2.27 -4.74
Intensity (A) of deprivation
among the poor (%)
49.1
51.9
51.2 -1.41 0.02 -0.62
Note: Calculations of authors’ using data sourced from NDHS, PSA.
*= 2014 approach to estimation of Global MPI
In 2017, the proportion in multidimensional poverty is estimated at 4.1 percent using the (old)
approach for the global MPI. This is just half of the World Bank’s estimate (8.3%) of the
proportion of Filipinos in consumption poverty who spend less than $1.9 in PPP 2011 prices10.
This estimate is a reduction of 4.7 percent per year in the period from 2008 to 2017. If we
consider instead the reduction of the adjusted headcount estimate, the rate of change is similar
at 5.1 percent. Both these rates of change are faster than the corresponding annual drops (3.7
percent, and 1.4 percent, respectively) in the World Bank estimate of consumption poverty
incidence in the Philippines and in the official income poverty headcount in the period from
2009 to 2015. While monetary poverty is technically not comparable to multidimensional
poverty from NDHS (using the MPI approach), it is interesting to note that estimates of
monetary headcount poverty are not decelerating as much as the estimates of headcounts for
multidimensional poverty for roughly the same periods.
The extremely poor, i.e., persons with half of the weighted deprivations, range from half to
two-thirds of the multidimensional headcount in the period from 2008 to 2017. Just like
headcount poverty, the proportion in severe poverty has reduced in the same period. Beyond
poverty, we can also look into distributional issues. The entire population may be broken down
into those in multidimensional poverty (who experience at least a third of total weighted
deprivations), those vulnerable to poverty, and those with no deprivations (Table 3). Across
time, those with no deprivations has been increasing from about 15 percent of the population
in 2008 to more than double this proportion nine years later. Further, the use of a lower cross-
dimensional cut-off of a fifth, rather than a third, increases the estimated poverty headcounts
by 53% to 66%.
10 See World Bank PovCalNet http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx
26
Table 3 Share of Population by Poverty and Vulnerability Status
Proportion of population 2017 2013 2008
in multidimensional poverty (who have higher than 33.32% intensity
of deprivations) 4.3 6.3 6.8
in severe poverty (with intensity higher than 50%) 2.3 4.2 4.5
who experience 20-33.32% intensity of deprivations 4.9 8.4 10.4
who experience more than 0 but less than 20% intensity of deprivations
58.7
68.2
67.6
with no deprivations (intensity=0) 32.1 17.1 15.2 Note: Calculations of authors’ using data sourced from NDHS, PSA.
*= 2014 approach to estimation of Global MPI
The contribution of the component dimensions of MPI for the global MPI approach is given in
Table 4. From 2008 to 2017, significant declines in the deprivation indicators related to
education, health and living standards dimension can be observed, though the improvement in
many living standards indicators, especially access to electricity and access to safe sanitation
appears to be more significant in the second sub-period 2013-17. Throughout the entire period
from 2008 to 2017, improvements in information assets (especially mobile phones) are
consistently observed. Overall, since most deprivation indicators are improving over time, we
expect multidimensional poverty to decline, as has been registered in Table 3, but clearly, how
the MPI reduces depends upon the changes in the joint distribution of the deprivation
indicators, as well as the weights for the component dimensions and indicators of the MPI.
Table 4 Incidence of Deprivation (in %) from the Global MPI Approach* : 2008-2017
Dimension Indicator/ Sub-Indicator 2008 2013 2017
Education E1: deprived in years of schooling 2.3 3.5 3.8
Health H1: deprived in child mortality 1.9 3.0 2.9
Living
Standards Deprived in Overall Living Standards 62.6 74.2 74.0
LS1: deprived in electricity 6.5 12.1 14.9
LS2: deprived in sanitation 21.8 29.5 30.6
LS3: deprived in source of water 7.0 9.6 10.1
LS4: deprived in floor quality 7.3 9.3 8.5
LS5: deprived in cooking fuel 49.4 62.7 64.8
LS6: deprived in assets 33.2 39.3 40.6
information asset-deprived (do not
own tv, radio, celphone, or landline) 2.9 5.7 9.5
livelihood-deprived 55.0 59.5 58.1
mobility asset-deprived (no
car/truck, motorcycle/tricycle,
bicycle) 44.7 52.0 53.2 Note: Authors’ calculations’ using data sourced from NDHS, PSA.
*= 2014 approach to estimation of Global MPI
In the next sections, we look further into trends of multidimensional poverty measures,
including the adjusted headcount (i.e. the MPI), resulting from the use of indicators listed in
Table 1 and compare patterns with those generated from the global MPI. We also carry out
examination of the robustness of underlying patterns generated if alternative weights to the
nested-equal weights approach, viz., based on PCA are used.
27
4.1. Trends
Trends in multidimensional poverty headcount estimates based on the indicators identified in
Table 1 are shown in Figure 6, together with estimates of monetary poverty headcounts in the
period 2008 to 2017. All the three surveys suggested a reduction in multidimensional poverty
headcount, but the levels of estimates of the headcount, and annual rates of change are survey-
dependent. APIS yielded a 10.0% reduction in multidimensional poverty headcount from
2016 to 2017; FIES had a 2.3% annual reduction from 2009 to 2015, while NDHS yielded a
4.6% reduction per year. The reduction in the FIES multidimensional poverty headcounts,
which are the least reduction across the three surveys, are midway that of the rates of reduction
suggested by World Bank poverty estimates and official poverty rates. If one were to expect
that economic growth should yield significant poverty reduction, it appears that the length of
the survey instruments in the FIES is likely seriously eroding the quality of monetary poverty
information and other aggregate information being generated (Albert et al. 2017), even non-
monetary indicators that are used for the MPI generation in this study. Except for 2012, the
FIES multidimensional poverty headcounts are fairly comparable to the World Bank estimates
of consumption poverty.
Figure 6 Multidimensional Poverty Headcount and Monetary Poverty Headcounts
Notes: (i) Multidimensional Poverty Headcounts are authors’ calculations’ using data sourced from NDHS,
PSA; (ii) World Bank consumption poverty rates were obtained from Povcalnet
http://iresearch.worldbank.org/PovcalNet/povOnDemand.aspx ; (iii) Official poverty headcount sourced from
PSA website https://psa.gov.ph/content/poverty-incidence-among-filipinos-registered-216-2015-psa .
28
Using the multidimensional poverty lens, the quality of life in the Philippines appears to be
consistently improving though the resulting estimates and reduction in estimates are not robust,
suggesting that the measurement crucially depends on the choice of indicators. Even when the
indicators are fairly comparable as in the case of the FIES and APIS, the results vary
considerably on account of the way the indicator is generated in the surveys. The expenditure
module in APIS is far simpler and shorter in length than the corresponding module in FIES.
However, it is interesting to notice that regardless of whether we look into monetary poverty
or multidimensional poverty aggregates, the reduction is mostly more evident in later years,
consistent also with results from the global MPI approach.
When traditional poverty measures, whether based on income- or consumption, are generated,
these are also presented through a poverty profile which describes the major facts on poverty
(Albert 2008; Haughton and Khander 2009; Albert et al. 2015). A poverty profile also
illustrates the pattern of poverty to see it varies by various subpopulations, such as geography
(urban or rural, regions, and so on), community characteristics (villages with and without a
public school), and household and individual characteristics (for example, educational
attainment of adult members or employment of household head). For instance, income poverty
by regions in recent years suggests, Metro Manila has had the lowest (income) poverty rate
across the regions at less than five percent of the population (followed by Calabarzon and
Central Luzon), while the Autonomous Region of Muslim Mindanao (ARMM) has the highest
(income) poverty headcount at more than half of the population (followed by Caraga, Eastern
Visayas and Soccsksargen).
Annex Table A-2 lists selected multidimensional poverty measures by region using the FIES
and APIS data. The multidimensional poverty headcounts based on FIES for the regions
correlate very strongly with the APIS-based measures, as well with the income poverty profiles
for the regions. Poverty is found to be worst in ARMM (at about a third of the population)
according to the FIES-based multidimensional poverty measurement, and least in Central
Luzon and Metro Manila, Calabarzon and Ilocos at under five percent. For the APIS-based
multidimensional poverty measures, ARMM still has the highest poverty headcounts while the
least headcounts are in Ilocos, Central Luzon, Calabarzon and Metro Manila.
Estimates of various poverty measures, viz., the official income poverty headcount, as well as
the multidimensional poverty headcount, and MPI as of 2015 are provided in Annex Table A-
3. These statistics, computed from 2015 FIES data, are given by economic sector of
employment of the household head, as well as by geographic area (region and urban/rural
location) for 2015, together with the contribution of the subpopulations to the poverty
measures. The poverty profiles are fairly similar when seen from any of the three poverty
measures. In both income- and multidimensional poverty measures, the largest concentration
of poverty is in agriculture (at least half of the poor population), in rural areas (about three
quarters of the population, and in Bicol, Western Visayas, Central Visayas, Soccsksargen, and
ARMM regions. While accounting for about 12.5% of the country’s population, Metro Manila
contributes only about 5% or less to total poverty. These empirical results are expected since
among subpopulations where the income-poor are concentrated, it is likely that the income
poor are also deprived of services, especially health and education, in part because of the
interlinkages of governance, geography and the provision of services.
As pointed out in earlier sections, the MPI, an adjusted headcount measure, is decomposable
for the dimensions of the index. The decomposition by dimensions shown in Table 6 allows
for an identification of the relative contribution of each dimension to the aggregate
29
multidimensional poverty measure. For the APIS, health is the standard of living dimension
that contributed the most to multidimensional poverty estimates, while for NDHS, education
had the biggest contribution. For FIES, living standards or education surpassed the
contribution of health. That the results are not robust is not surprising as indicators are not
available across all the three surveys. The seemingly most comparable set of indicators are
from the FIES and APIS. However, since expenditure is of less detail in the APIS, the food
expenditure-based indicator (which proxies health and nutrition) may have had much less
variability in the FIES than the corresponding one in the APIS. Consequently, this may have
had led to the varying results on the contributions of the dimensions to MPI estimates across
surveys.
Table 5 Contribution to MPI by Dimension
Data Source MPI Percent Contribution of Dimension to MPI
Education Health Living
Standards
2017 APIS 0.028 20.8 46.8 32.4
2016 APIS 0.039 22.0 45.8 32.1
2014 APIS 0.059 19.7 46.8 33.5
2015 FIES 0.045 47.4 1.9 50.7
2012 FIES 0.057 47.4 1.4 51.2
2009 FIES 0.047 52.0 0.7 47.3
2017 NDHS 0.019 40.2 33.3 27.3
2013 NDHS 0.030 38.0 33.1 29.6
2008 NDHS 0.031 38.7 31.6 30.1 Notes: Authors’ calculations’ using data sourced from APIS, FIES, and NDHS, PSA
While there should be correlation between multidimensional poverty and monetary poverty,
the relationship is not expected to be one-to-one. Some of those who are not deprived in
monetary terms may have other deprivations, and those who may have deprivations in some
monetary dimensions (such as education or health) need not be poor in income or consumption.
However, we would expect that a substantial proportion of Filipinos who are from the lower
part of (per capita) income distribution should be considered MPI poor or MPI vulnerable.
Further, a substantial share of the upper part of income distribution should be without
deprivations.
Table 7 gives a breakdown of Filipinos by (per capita) income clusters (see Albert et al. 2018)
and by multidimensional poverty or vulnerability status in 2015 according to the FIES. Among
the income poor, only a fifth (20.3%) are MPI-poor, but three-quarters (76.7%) are MPI-
vulnerable and the remaining 3.0% are without deprivations. While among those who are low-
income but not poor11, nine-tenths (89.4%) are either MPI-poor or MPI-vulnerable, and the
remaining proportion (10.6%) are found to be without deprivations. Of the non-lower income
group (which comprises about two -fifths of all Filipinos), one-twentieth (6.%) have at least
20% deprivations (either MPI-poor or MPI-vulnerable), although the bulk of this is among the
lower middle-income cluster.
11 Those with per capita income higher than the poverty threshold, but lower than twice the poverty threshold (as per Albert et al. 2018).
30
Table 6 Distribution of Filipinos by (Per capita) Income Cluster and by MPI-Poverty or Vulnerability Status: 2015
Income Cluster MPI poor MPI-
vulnerable
with 20.0% -
33.3%
deprivations
MPI-
vulnerable
with 0%-
20.0%
deprivations
Not MPI
vulnerable,
i.e. with 0%
deprivations
Total
Poor 4.4 5.3 11.3 0.7 21.6
Low income but not poor 2.3 5.9 24.6 3.9 36.8
Lower middle income 0.4 1.8 17.2 7.0 26.4
Middle income 0.0 0.2 5.5 4.5 10.2
Upper middle income 0.0 0.1 1.3 2.3 3.7
Upper income but not rich 0.0 0.0 0.3 0.8 1.1
Rich 0.0 0.0 0.1 0.3 0.4
Total 7.2 13.2 60.3 19.3 100.0
Notes: Authors’ calculations’ using data sourced from FIES, PSA
Similarly, when we examine the APIS data, we would expect few of the poorest in income
distribution to be viewed as not having any multidimensional deprivations, and likewise have
few MPI-poor among those from the topmost of per capita income distribution. Based on the
2017 APIS, we find only 6.0 percent of those in the bottom (per capita) income quintile being
considered MPI-not vulnerable (Table 8). Further, among the top three quintiles, only a tenth
(9.8%) are found to have at least 20% deprivations (i.e. being MPI poor or MPI-vulnerable
with 20-33 percent deprivations).
Table 7 Distribution of Filipino households by (Per capita) Income Quintile and by MPI-Poverty or Vulnerability Status: 2017
Income Quintile MPI poor MPI-
vulnerable
with 20.0% -
33.3%
deprivations
MPI-
vulnerable
with 0%-
20.0%
deprivations
Not MPI
vulnerable,
i.e. with 0%
deprivations
Total
Lowest quintile 4.1 3.3 11.4 1.2 20.0
Second lowest quintile 1.7 2.0 13.1 3.2 20.0
Middle quintile 1.4 1.3 12.8 4.5 20.0
Second to the richest
quintile 1.5 0.6 11.8 6.1 20.1
Richest quintile 0.9 0.2 9.0 9.9 19.9
Total 9.5 7.4 58.2 25.0 100.0
Notes: Authors’ calculations’ using data sourced from APIS, PSA
4.2. Robustness
We further explore the issue of robustness of the empirical results by looking into two
weighting schemes discussed earlier (the nested equal weights and weights obtained from
principal components analysis). The results are shown in Table 9 for the indicators across the
three waves of surveys employed in this study. Both weighting schemes examined here are
fixed over time. Notable differences in estimates of multidimensional poverty measures are
observed across the two sets of weights (as in the empirical findings of Datt 2017, who also
tried out alternative weighting schemes for MPI estimation). We notice that the measures using
31
PCA-based weights are much higher (twice to triple) compared to those using nested equal
weights for both the FIES and APIS. Further, for the FIES data, the multidimensional poverty
headcounts using PCA-based weights are fairly similar (and, in fact, even slightly higher)
estimates of the proportion in poverty, compared to the official poverty headcounts. The PCA-
based multidimensional poverty headcount estimates dropped only by about one percentage
point from 2009 to 2012, while the estimates from 2012 to 2015 reduced by four percentage
points. Interestingly, this reduction is similar also to the decreases (in percentage points) for
the official (income-based) poverty headcounts during the same period. The lack of
robustness in use of composite indices confirms typical expectations and findings that such
indices are crucially dependent upon the choice of indicators, and the weights used for these
indicators, and put serious question on whether we may be actually observing the precise
quantity of signals in multidimensional poverty changes across time.
32
Table 8 Multidimensional Poverty Measures Using Nested Equal Weights (NEW) and Principal Components Analysis (PCA)-Based Weights: 2008-2017 (a) FIES
Multidimensional
Poverty Measure
2015 FIES 2012 FIES 2009 FIES
NEW PCA NEW PCA NEW PCA
Multidimensional
Poverty Index
(MPI = H*A)
0.028
0.126
0.039
0.149
0.059
0.160
Headcount ratio: H
Population in
multidimensional
poverty (%)
7.2
25.1
9.7
28.9
14.0
29.9
Intensity A of
deprivation among
the poor (%)
39.6
50.1
40.1
51.7
42.3
53.5
Proportion (in %)
multidimensionally
vulnerable to
poverty (who
experience more
than 0 but less than
20% intensity of
deprivations)
60.3
34.3
59.1
34.0
53.1
33.1
Proportion (in %)
multidimensionally
vulnerable to
poverty (who
experience 20-
33.32% intensity of
deprivations)
13.2
21.3
15.9
21.8
18.1
22.2
Proportion (in %) of
the population with
no deprivations
(intensity=0)
19.3
19.3
15.4
15.4
14.9
14.9
Proportion (in %) of
the population in
severe poverty (with
intensity higher than
50%)
1.0
10.5
1.4
13.4
3.5
14.7
Contribution of
education
dimension to
multidimensionally
poor (in %)
47.4
9.9
47.4
10.8
52.0
17.1
Contribution of
health dimension to
multidimensionally
poor (in %)
1.9
0.0
1.4
0.0
0.7
0.0
Contribution of
living standards
dimension to
multidimensionally
poor (in %)
50.7
90.0
51.2
89.1
47.3
82.9
total number of
sampled households
41,539
41,539
40,168
40,168
38,400
38,400
Notes: Authors’ calculations’ using data sourced from FIES, PSA
33
(b) APIS Multidimensional
Poverty Measure 2017 APIS 2016 APIS 2014 APIS
NEW PCA NEW PCA NEW PCA
Multidimensional
Poverty Index
(MPI = H*A)
0.045
0.086
0.057
0.106
0.047
0.112
Headcount ratio: H
Population in
multidimensional poverty
(%)
9.5
18.0
11.9
21.3
9.9
23.1
Intensity A of
deprivation among the
poor (%)
47.4
47.9
47.8
49.7
47.2
48.5
Proportion (in %)
multidimensionally
vulnerable to poverty
(who experience more
than 0 but less than 20%
intensity of deprivations)
58.2
38.3
58.1
38.9
59.0
35.9
Proportion (in %)
multidimensionally
vulnerable to poverty
(who experience 20-
33.32% intensity of
deprivations)
7.4
18.8
9.0
18.7
9.0
18.9
Proportion (in %) of the
population with no
deprivations
(intensity=0)
25.0
25.0
21.1
21.1
22.1
22.1
Proportion (in %) of the
population in severe
poverty (with intensity
higher than 50%)
3.7
6.5
4.9
9.1
4.0
9.2
Contribution of
education dimension to
multidimensionally poor
(in %)
20.8
7.8
22.0
8.0
19.7
5.9
Contribution of health
dimension to
multidimensionally poor
(in %)
46.8
5.2
45.8
5.0
46.8
4.6
Contribution of living
standards dimension to
multidimensionally poor
(in %)
32.4
87.0
32.1
86.9
33.5
89.5
total number of sampled
households
9,732
9,732
10,295
10,295
7,831
7,831
Notes: Authors’ calculations’ using data sourced from APIS, PSA
34
5. Summary and Policy Implications
With income poverty reduction being meager in the Philippines in recent years despite strong
economic growth suggests that that economic growth has not been inclusive (Albert et al 2014).
This study shows that multidimensional measures of poverty indicate a far bigger decline in
poverty than what can be observed from income or consumption poverty metrics, whether seen
from official statistics or the poverty rates estimated using international poverty lines.
However, the multidimensional measures are not robust, both in levels, in the reductions, and
in the contributions of the dimensions. The empirical results in the estimation fundamentally
depend on the choice of the data source, the component indicators used, as well as the selection
of weights for the indicators.
There is unanimity in recognition that poverty is multidimensional, and that having a single
indicator of poverty, whether based on income or consumption data, will in no way fully
capture the complexity of poverty. Having a single composite index for summarizing
multidimensional poverty may seem attractive, as in the case in measuring progress and quality
of life, but it is unclear how such a composite index for multidimensional poverty can really
contribute to better thinking about poverty, or better policies for eradicating poverty.
The MPI component indicators are a combination of data on stocks and flows, and of inputs to
economic well-being and social development outcomes, which makes the composite index
appear like a fruit salad that combines apples, oranges, grapes, and other fruits. The global MPI
and its variants do not put a dimension on labor and employment, although there is nothing that
stops analysts from using a different set of dimensions (see, e.g. Datt 2017). When generating
composite indices, the fundamental question are (a) whether the entire exercise adds “apples
and oranges” ; (b) if so, what fruits should be mixed? (c) and in what proportions? In measuring
income or consumption poverty, an analyst relies on economic theory, which says that under
certain conditions market prices provide the correct weights for the aggregation of the monetary
indicator. For composite indices such as the MPI, HDI and SPI, no consensus exists on how
the multiple dimensions and indicators should be weighted to form the index. That these
indices add up fundamentally different things in a rather arbitrary way suggests this may not
necessarily be the way to refine poverty measurement even if current methodology is imperfect
especially in capturing the multidimensional nature of poverty.
While we have to recognize that there are important aspects of welfare that cannot be captured
in the proportion of persons with consumption less than $1.90 per person per day (in PPP 2011
prices), or even with the official income poverty rates released by NSOs such as the PSA, but
neither can everything be put into a single index, whether the MPI, or even the HDI or SPI.
These composite indices can certainly be used for policy advocacy to show disparities across
countries, or disparities within countries, but ultimately, policies will have to examine the
specific components of these indices.
Government is well advised to tread carefully in its decision to generate an official measure of
multidimensional poverty. Should it continue with its decision, the PSA and NEDA should
work out a communication plan for explaining a change in the indices to be generated, and how
this measurement system ultimately relates with the official income poverty measurement. If
the communication strategy will merely attempt to show changes in the components of the
indices, then this may not be helpful.
35
The generation of multidimensional measures of poverty may, however, be justified from the
perspective that poverty is not static, and neither should its measurement, especially given the
various risks to future poverty that people face, and the intersections of the various dimensions
of poverty with traditional poverty measurement(see Table 8). Toward this end, if government
decides to start working on a multidimensional poverty measure, it is important to decide the
specific data source for the actual measurement. Given that APIS already makes use largely of
the FIES income schedule, and that PSA generates poverty statistics from the half semester
income data of the APIS, there may be some opportunities of exploring the wealth of welfare
indicators in the APIS to relate monetary poverty with multidimensional poverty. The use of
indicators, such as experience of hunger, and non-coverage of health insurance, other than
those used in this study, may also be looked into. The estimation of multidimensional poverty,
however, might be best left not to the PSA itself, but to research institutions in order not to
confuse the public about the different estimates of multidimensional poverty headcounts, and
official income poverty headcounts. Clearly, this study suggests that multidimensional poverty
estimation has linkages with income (and consumption) poverty, but the results are not robust.
Further, this study also suggests that economic growth in recent years has translated to better
quality of life, both in income and non-monetary measures and that government will need to
find ways of not only ensuring inclusive prosperity amidst growth, as well as examine prospects
for improving current understanding of the many dimensions of poverty so that as we continue
our economic growth path, no person will be left behind.
36
6. References
Albert, J. R. G. 2008. Issues on counting the poor. Policy Notes No. 2008-11. Philippine
Institute for Development Studies (Available at
https://dirp4.pids.gov.ph/ris/pn/pidspn0811.pdf ; Accessed 15 October 2018)
Albert, J. R. G., Santos, A. G. F., and Vizmanos, J. F. V. 2018. Defining and Profiling the
Middle Class. Policy Notes (to be released). Philippine Institute for Development
Studies
Albert, J. R. G., Asis, R. D., and Vizmanos, J. F. V. 2017. Why differences in household
expenditure estimates matter. Policy Notes No. 2015-06. Philippine Institute for
Development Studies (Available at
https://pidswebs.pids.gov.ph/CDN/PUBLICATIONS/pidspn1706.pdf ; Accessed 6
December 2018)
Albert, J. R. G., Dumagan, J. C. and Martinez, A., Jr. 2015. Inequalities in Income, Labor, and
Education: The Challenge of Inclusive Growth. Discussion Paper No. 2015-01.
Philippine Institute for Development Studies (Available at
https://dirp3.pids.gov.ph/webportal/CDN/PUBLICATIONS/pidsdps1501.pdf ;
Accessed 15 October 2018)
Albert, J. R. G., and Molano, W. 2009. Estimation of the Food Poverty Line. Discussion Paper
No. 2009-14. Philippine Institute for Development Studies (Available at
https://dirp4.pids.gov.ph/ris/dps/pidsdps0914.pdf ; Accessed 15 October 2018)
Alkire, S. and Foster, J. 2011. Counting and multidimensional poverty measurement. Journal
of Public Economics, vol. 95(7–8), pp. 476–487. (Available at
https://www.ophi.org.uk/wp-content/uploads/ophi-wp7.pdf ; Accessed 15 October
2018)
Alkire, S., Kanagaratnam, U., and Suppa, N. 2018. The Global Multidimensional Poverty Index
(MPI): 2018 Revision. (Available at https://ophi.org.uk/wp-
content/uploads/OPHI_MPI_Meth_Note_46.pdf ; Accessed 15 November 2018)
Atkinson, A. 2003. Multidimensional Deprivation: Contrasting Social Welfare and Counting
Approaches, Journal of Economic Inequality, 1(1): 51–65. 91 (Available at
https://pdfs.semanticscholar.org/462f/f6306155764fb76a04b440399b0a54322e00.pdf;
Accessed 15 October 2018)
Birdsall, N. (2011) Comment on Multi-dimensional Indices. Journal of Economic Inequality
9(3): 489–91 (Available at https://link.springer.com/content/pdf/10.1007%2Fs10888-
011-9195-y.pdf ; Accessed 15 November 2018)
Bourguignon, F., and S. Chakravarty. 2003. The Measurement of Multidimensional
Poverty.”Journal of Economic Inequality 1 (1): 25–49. (Available at
http://www.ophi.org.uk/wp-content/uploads/Bourgignon-Chakravarty-2003.pdf ;
Accessed 15 November 2018)
37
Centre for Bhutan Studies (CBS) and GNH Research. 2015. A Compass Towards A Just and
Harmonious Society: 2015 GNH Survey Report. (Available at
http://www.bhutanstudies.org.bt/wp-content/uploads/2017/05/2015-Survey-
Results.pdf ; Accessed 20 October 2018)
Commission on Growth and Development (CGD) 2008. The Growth Report: Strategies for
Sustained Growth and Inclusive Development. (Available at
http://siteresources.worldbank.org/EXTPREMNET/Resources/489960-
1338997241035/Growth_Commission_Final_Report.pdf; ; Accessed 12 November
2018)
Ericta, C. N., and Fabian, E. 2009. A Documentation of the Philippines’ Family Income and
Expenditure Survey. Discussion Paper No. 2009-18. Philippine Institute for
Development Studies (Available at https://dirp4.pids.gov.ph/ris/dps/pidsdps0918.pdf ;
Accessed 15 October 2018)
Ericta, C. N., and Luis, J. 2009. A Documentation of the Annual Poverty Indicators Survey.
Discussion Paper No. 2009-20. Philippine Institute for Development Studies (Available
at https://dirp3.pids.gov.ph/ris/dps/pidsdps0920.pdf ; Accessed 15 October 2018)
Ferreira, F.H.G. and M.A. Lugo (2013) ‘Multidimensional Poverty Analysis: Looking for a
Middle Ground’, The World Bank Research Observer 28(2): 220–35. World Bank,
Washington, DC (Available at
https://openknowledge.worldbank.org/bitstream/handle/10986/21430/wbro_28_2_220
.pdf ; Accessed 15 November 2018)
Filmer, D., & Pritchett, L. 2001. Estimating Wealth Effects without Expenditure Data-or Tears:
An Application to Educational Enrollments in States of India. Demography, 38(1), 115-
132. (Available at http://www.jstor.org/stable/3088292 ; Accessed 12 November 2018)
Graham, C. 2009. Happiness Around the World: The Paradox of Happy Peasants and Miserable
Millionaires. Oxford University Press Inc., New York.
Gwatkin, D., Rutstein, S., Johnson, K., Pande, R., and Wagstaff, A. 2000. Socio-economic
Differences in Health, Nutrition, and Population in Bangladesh. (and comparable
publications covering Benin, Bolivia, Brazil, Burkina Faso, Cameroun, Central African
Republic, Colombia, Comores, Côte d’Ivoire, Dominican Republic, Ghana, Guatemala,
Haiti, India, Indonesia, Kenya, Kyrgyz Republic, Madagascar, Malawi, Mali, Morocco,
Mozambique, Namibia, Nepal, Nicaragua, Niger, Nigeria, Pakistan, Paraguay, Peru,
Philippines, Senegal, 3 Tanzania, Togo, Turkey, Uganda, Vietnam, Zambia, and
Zimbabwe.) World Bank, Washington, DC.
Haughton, J. H. and Khandker, S. 2009. Handbook on poverty and inequality. (Available at
https://openknowledge.worldbank.org/bitstream/handle/10986/11985/9780821376133
.pdf ; Accessed 12 November 2018)
Kraay, A. 2018. Methodology for a World Bank Human Capital Index. Policy Research
Working Paper 8593. World Bank, Washington, DC (Available at
http://documents.worldbank.org/curated/en/300071537907028892/pdf/WPS8593.pdf ;
Accessed 15 November 2018)
38
Kovacevic, Milorad and Cecilia Calderon, M. 2014. UNDP’s Multidimensional Poverty Index:
2014 Specifications. UNDP Human Development Report Office Occasional Paper.
(Available at
http://hdr.undp.org/sites/default/files/specifications_for_computation_of_the_mpi_0.p
df ; Accessed 12 November 2018)
National Economic Development Authority. 2015. Ambisyon 2040: A Long Term Vision for
the Philippines. (Available at http://2040.neda.gov.ph/wp-content/uploads/2016/04/A-
Long-Term-Vision-for-the-Philippines.pdf ; Accessed 15 October 2018)
Organization for Economic Cooperation and Development (OECD) 2015. How's
Life?: Measuring Well-being. Paris: OECD Publishing. (Available at
https://doi.org/10.1787/how_life-2015-en.; Accessed 20 October 2018)
OECD 2011. How's Life?: Measuring Well-being. Paris: OECD Publishing. (Available at
https://doi.org/10.1787/9789264121164-en ; Accessed 20 October 2018)
OECD 2001. The DAC Journal 2000: Sweden, Switzerland Volume 1 Issue 4, OECD
Publishing, Paris. (Available at https://doi.org/10.1787/journal_dev-v1-4-en ;
Accessed 15 November 2018)
Pezzey, J. 1992. Sustainable Development Concepts: An Economic Analysis. World Bank
Environment Paper No. 2. World Bank, Washington DC. (Available at
http://documents.worldbank.org/curated/en/237241468766168949/pdf/multi-
page.pdf; Accessed 15 November 2018)
Ravallion, M. 2011. On multidimensional indices of poverty. Journal of Economic Inequality
9, 235-248. (Available at
http://cgdev.org.488elwb02.blackmesh.com/doc/event%20docs/Multidimensional-
Indices-of-Poverty.pdf ; Accessed 15 November 2018)
Ravallion, M. 2010. Troubling Tradeoffs in the Human Development Index. Policy Research
Working Paper 5484. World Bank, Washington DC (Available at
http://documents.worldbank.org/curated/en/302501468167375895/pdf/WPS5484.pdf ;
Accessed 10 October 2018)
Ravallion, M. 2012. Mashup indices of development. World Bank Research Observer 27(1),
1-32. World Bank, Washington DC (Available at
http://documents.worldbank.org/curated/en/454791468329342000/pdf/WPS5432.pdf ;
Accessed 15 November 2018)
Social Weather Stations (SWS). 2018. Press Release: Fourth Quarter 2017 Social Weather
Survey: Record-high 94% of Pinoys are “Very/Fairly Happy”; Record-high 92% are
“Very/Fairly Satisfied” with Life. (Available at
https://www.sws.org.ph/swsmain/artcldisppage/?artcsyscode=ART-20180320191125;
Accessed 12 November 2018)
Soubbotina, T. P. 2004. Beyond Economic Growth : An Introduction to Sustainable
Development, Second Edition. Washington, DC: World Bank. (Available at
39
http://documents.worldbank.org/curated/en/454041468780615049/pdf/2489402nd0ed
ition0Beyond0economic0growth.pdf ; Accessed 15 October 2018)
Stiglitz-Sen-Fitoussi Commission. 2009. Report by the Commission on the Measurement of
Economic Performance and Social Progress. (Available at
https://ec.europa.eu/eurostat/documents/118025/118123/Fitoussi+Commission+report
; Accessed 12 November 2018)
Sustainable Development Solutions Network (SDSN) 2018. World Happiness Report.
(Available at https://s3.amazonaws.com/happiness-report/2018/WHR_web.pdf ; ;
Accessed 12 November 2018)
United Nations (UN). The Millennium Development Goals Report 2015a. New York: UN
(Available at
http://www.un.org/millenniumgoals/2015_MDG_Report/pdf/MDG%202015%20rev
%20(July%201).pdf ; Accessed 15 October 2018)
UN. The Millennium Development Goals Report 2015a. New York: UN (Available at
http://www.un.org/millenniumgoals/2015_MDG_Report/pdf/MDG%202015%20rev
%20(July%201).pdf ; Accessed 15 October 2018)
UN Development Programme (UNDP) 2010. Human Development Report: The Real Wealth
of Nations: Pathways to Human Development, New York, New York, USA: UNDP
(Available at
http://hdr.undp.org/sites/default/files/reports/270/hdr_2010_en_complete_reprint.pdf ;
Accessed 15 October 2018)
UNDP 1990. Human Development Report. New York, New York, USA: UNDP. (Available at
http://hdr.undp.org/sites/default/files/reports/219/hdr_1990_en_complete_nostats.pdf ;
Accessed 15 October 2018)
UN, OECD and the Statistical Office for European Communities (Eurostat). 2009. Measuring
Sustainable Development. New York and Geneva. (Available at
https://www.oecd.org/greengrowth/41414440.pdf ; Accessed 12 November 2018)
UN Statistics Division (UNSD). 2005. Handbook on Poverty Statistics: Concepts, Methods
and Policy Use. New York: UN (Available at
https://unstats.un.org/unsd/methods/poverty/pdf/un_book%20final%2030%20dec%20
05.pdf ; Accessed 15 October 2018)
Ura, K., Alkire S., Zangmo, T., and Wangdi, K. 2012. An Extensive Analysis of GNH Index.
Centre for Bhutan Studies. (Available at http://www.grossnationalhappiness.com/wp-
content/uploads/2012/10/An%20Extensive%20Analysis%20of%20GNH%20Index.pd
f ; Accessed 20 October 2018)
Virola, R. A. 2011. Refinements on the Official Poverty Estimation Methodology and the
Sources of Differences of the Official Poverty Statistics and the NHTS-PR Estimates.
Meeting of the Cabinet Cluster on Human Development and Poverty Reduction. 1
December 2011. (Available at
40
https://psa.gov.ph/sites/default/files/Poverty_Cabinet_Cluster_01Dec2011.pdf ;
Accessed 10 October 2018)
Virola, R. A. and Encarnacion, J. O. 2007. Measuring Progress of Philippine Society: Gross
National Product or Gross National Happiness. Proceedings of the 10th National
Convention on Statistics. 1-2 October 2007. (Available at
http://nap.psa.gov.ph/ncs/10thNCS/papers/invited%20papers/ips-28/ips28-03.pdf ;
Accessed 12 November 2018)
Wolfers, J., Stevenson, B. 2008. Economic Growth and Subjective Well-Being: Reassessing
the Easterlin Paradox. Brookings Institution (Available at
http://users.nber.org/~jwolfers/papers/EasterlinParadox.pdf ; Accessed 12 November
2018)
41
Annex Table A-1 Dimensions and Indicators of the Multidimensional Poverty Index.
Dimension Indicators (and weights) 2018 2014 2010 Education a) Years of schooling (1/6)
Deprived if no household
member aged ten years or
older has completed six years of
schooling.
Deprived if no household
member has completed five
years of schooling.
Deprived if no household
member has completed five
years of schooling.
b) Child school attendance (1/6) Deprived if any school-aged
child is not attending school up
to the age at which he/she would
complete class 8a.
Deprived if any school-aged
child is not attending school up
to the age at which he/she would
complete class 8a.
Deprived if any school-aged
child is not attending school up
to the age at which he/she would
complete class 8a.
Health a) Child mortality (1/6) Deprived if any child has died in
the family in the five-year
period preceding the survey.
Deprived if any child has died in
the family.
Deprived if any child has died in
the family.
b) Nutrition (1/6): Deprived if any person under 70
years of age for whom there is
nutritional information is
undernourished b2.
Deprived if any adult or child for
whom there is nutritional
information is malnourished b1.
Deprived if any adult or child for
whom there is nutritional
information is malnourished b1.
Standard of Living a) Electricity (1/18) Deprived if the household has no
electricity.
Deprived if the household has no
electricity.
Deprived if the household has no
electricity.
b) Sanitation (1/18)
Deprived if the household’s
sanitation facility is not
improved (according to SDG
guidelines) or it is improved but
shared with other households. (A
household is considered to have
access to improved sanitation if
it has some type of flush toilet or
latrine, or ventilated improved
pit or composting toilet,
provided that they are not
shared. If survey report uses
other definitions of ‘adequate’
sanitation, we follow the survey
report.)
Deprived if the household’s
sanitation facility is not
improved (according to MDG
guidelines), or it is improved but
shared with other households. (A
household is considered to have
access to improved sanitation if
it has some type of flush toilet or
latrine, or ventilated
improved pit or composting
toilet, provided that they are not
shared.)
Deprived if the household’s
sanitation facility is not
improved (according to MDG
guidelines), or it is improved but
shared with other households. (A
household is considered to have
access to improved sanitation if
it has some type of flush toilet or
latrine, or ventilated
improved pit or composting
toilet, provided that they are not
shared.)
42
Dimension Indicators (and weights) 2018 2014 2010 c) Safe Drinking water (1/18) Deprived if the household does
not have access to improved
drinking water (according to
SDG guidelines) or safe
drinking water is at least a 30-
minute walk from home,
roundtrip. (A household has
access to clean drinking water if
the water source is any of the
following types: piped water,
public tap, borehole or pump,
protected well, protected spring
or rainwater, and it is within a
30-minute walk (round trip). If
survey report uses other
definitions of ‘safe’ drinking
water, we follow the survey
report.)
Deprived if the household does
not have access to safe drinking
water (according to MDG
guidelines) or safe drinking
water is more than a 30-minute
walk from home roundtrip. (A
household has access to clean
drinking water if the water
source is any of the following
types: piped water, public tap,
borehole or pump, protected
well, protected spring or
rainwater, and it is within a
distance of 30 minutes’ walk
(roundtrip).)
Deprived if the household does
not have access to safe drinking
water (according to MDG
guidelines) or safe drinking
water is more than a 30-minute
walk from home roundtrip. (A
household has access to clean
drinking water if the water
source is any of the following
types: piped water, public tap,
borehole or pump, protected
well, protected spring or
rainwater, and it is within a
distance of 30 minutes’ walk
(roundtrip).)
d) Floor (1/18) Deprived if the household has
inadequate housing: the floor is
of natural materials or the roof
or wall are of rudimentary
materials. (i.e., if floor is made
of mud/clay/earth, sand or dung;
or if dwelling has no roof or
walls or if either the roof or
walls are constructed using
natural materials such as cane,
palm/trunks, sod/mud, dirt,
grass/reeds, thatch, bamboo,
sticks, or rudimentary materials
such as carton, plastic/ polythene
sheeting, bamboo with
mud/stone with mud, loosely
packed stones, uncovered adobe,
raw/reused wood, plywood,
Deprived if the household has a
dirt, sand, or dung floor.
Deprived if the household has a
dirt, sand, or dung floor.
43
Dimension Indicators (and weights) 2018 2014 2010 cardboard, unburnt brick or
canvas/tent.)
e) Cooking fuel (1/18) Deprived if the household cooks
with dung, wood or charcoal.
Deprived if the household cooks
with dung, wood or charcoal.
Deprived if the household cooks
with dung, wood or charcoal.
f) Assets ownership (1/18) Deprived if the household does
not own more than one of these
assets: radio, TV, telephone,
computer, animal cart, bicycle,
motorbike or refrigerator, and
does not own a car or truck.
Deprived if the household does
not own more than one radio,
TV, telephone, bike, motorbike
or refrigerator and does not own
a car or truck.
Deprived if the household does
not own more than one radio,
TV, telephone, bike, motorbike
or refrigerator and does not own
a car or truck.
Notes: a = Data source for age children start primary school: United Nations Educational, Scientific and Cultural Organization, Institute for Statistics database, Table 1. Education systems b1 = Adults are considered malnourished if their BMI is below 18.5 m/kg2. Children are considered malnourished if their z-score of weight-for-age is below minus two standard deviations. b2 = Adults 20 to 70 years are considered malnourished if their Body Mass Index (BMI) is below 18.5 m/kg2. Those 5 to 20 are identified as malnourished if their age-specific BMI cutoff is below
minus two standard deviations. Children under 5 years are considered malnourished if their z-score of either height-for-age (stunting) or weight-for-age (underweight) is below minus two standard
deviations from the median of the World Health Organization 2006 reference population. In a majority of the countries, BMI-for-age covered people aged 15 to19 years, as anthropometric data
was only available for this age group; if other data were available, BMI-for-age was applied for all individuals above 5 years and under 20 years.
44
Annex Table A-2 Intensity of Deprivation, Multidimensional Poverty Headcount and Proportion of Population Deprived in Living
Standards, Education and Health Dimensions : 2009-2017
(a) FIES
Region FIES 2015 FIES 2012 FIES 2009
intensity headcount living
standards
deprived
education
deprived
health
deprived
intensity headcount living
standards
deprived
education
deprived
health
deprived
intensity headcount living
standards
deprived
education
deprived
health
deprived
Ilocos Region 35.5 3.6 74.4 15.4 0.1 36.7 3.7 77.4 16.1 0.3 38.8 5.8 80.7 22.4 0.1
Cagayan Valley 38.5 4.7 68.9 15.3 0.3 37.7 6.9 81.2 18.2 0.2 39.5 10.8 77.8 30.9 0.1
Central Luzon 36.0 2.1 64.3 14.0 0.1 36.9 3.7 70.8 19.1 0.1 40.7 4.3 69.7 25.2 0.0
Bicol Region 38.3 8.7 84.8 18.5 0.1 39.1 10.4 87.4 21.2 0.1 42.8 19.6 87.3 38.2 0.1
Western Visayas 40.1 9.8 86.9 16.1 0.1 40.8 11.7 87.7 18.9 0.1 42.4 20.3 89.9 31.6 0.0
Central Visayas 39.3 8.1 80.7 19.0 0.4 40.1 12.6 84.8 21.6 0.3 43.0 20.8 84.2 34.6 0.3
Eastern Visayas 39.1 9.2 83.6 20.8 0.2 38.7 15.2 87.8 26.7 0.2 41.8 21.4 85.3 40.3 0.2
Western
Mindanao
40.7 12.4 84.7 21.6 0.5 42.2 19.0 87.7 27.7 0.7 43.6 29.7 87.8 44.3 0.5
Northern
Mindanao
39.6 7.7 82.5 18.3 0.3 40.8 9.8 82.4 24.9 0.3 41.1 16.2 84.5 35.0 0.2
Southern
Mindanao
39.7 8.0 77.5 18.4 0.1 41.2 12.0 82.3 24.0 0.1 43.6 17.7 83.1 36.5 0.5
Central
Mindanao
40.3 15.1 84.4 22.9 0.1 41.1 18.5 88.1 28.3 0.3 43.6 20.2 86.7 35.3 0.1
NCR 36.8 2.4 78.6 11.0 0.1 36.1 3.2 81.7 14.8 0.0 36.9 3.2 82.2 18.6 0.0
CAR 38.7 3.9 82.3 13.9 0.2 37.8 4.9 86.5 18.1 0.1 39.7 8.6 85.7 25.6 0.1
ARMM 41.8 29.3 97.8 35.8 0.0 41.6 37.2 98.5 43.8 0.1 43.3 40.6 96.7 53.7 0.0
CARAGA 39.6 7.3 78.5 20.8 0.4 39.7 10.9 87.2 24.2 0.2 41.1 15.6 87.8 32.8 0.2
CALABARZON 37.2 3.2 76.8 13.3 0.1 37.5 3.8 82.0 18.4 0.1 40.3 6.7 79.9 24.5 0.0
MIMAROPA 42.4 11.4 79.7 20.2 0.2 42.8 17.0 84.6 24.0 0.2 44.8 23.5 85.7 37.8 0.1
PHILIPPINES 39.6 7.2 79.0 17.0 0.2 40.1 9.7 83.1 21.3 0.2 42.3 14.0 82.9 30.8 0.1
Notes: Authors’ calculations’ using data sourced from FIES, PSA
45
(b) APIS Region APIS 2017 APIS 2016 APIS 2014
intensity headcount living
standards
deprived
education
deprived
health
deprived
intensity headcount living
standards
deprived
education
deprived
health
deprived
intensity headcount living
standards
deprived
education
deprived
health
deprived
Ilocos Region 40.0 4.9 63.3 6.5 4.3 41.6 5.7 67.2 7.6 3.7 42.0 17.3 73.1 10.3 3.9
Cagayan Valley 42.8 11.7 55.6 14.9 5.1 42.5 13.0 61.2 10.2 8.1 43.4 12.2 59.4 9.1 7.5
Central Luzon 45.5 5.2 58.1 6.4 1.8 44.8 5.0 64.8 7.5 5.5 40.4 7.4 65.4 9.2 2.9
Bicol Region 45.8 8.6 67.4 7.5 4.9 44.6 8.9 76.5 13.0 4.0 48.7 37.3 84.0 13.4 5.6
Western Visayas 48.5 23.6 80.1 12.1 6.0 51.3 34.4 85.0 12.9 7.0 50.3 28.2 83.4 12.0 7.5
Central Visayas 53.2 31.2 85.2 16.0 6.4 48.9 31.3 84.2 11.1 8.6 51.6 34.6 75.2 12.5 12.0
Eastern Visayas 47.4 23.8 79.9 7.9 11.7 50.1 31.6 81.6 13.8 14.9 49.8 40.9 86.3 15.1 15.9
Western
Mindanao 47.3 27.7 84.2 13.5 7.7 46.3 37.0 84.0 18.7 10.2 47.8 36.7 80.5 10.7 11.8
Northern
Mindanao 49.7 35.1 79.3 16.2 10.4 50.3 36.5 89.0 14.3 14.4 49.7 36.5 78.8 10.9 19.9
Southern
Mindanao 47.2 30.0 84.9 11.9 10.6 49.0 24.8 76.3 13.1 13.7 48.9 27.6 82.7 10.3 10.5
Central
Mindanao 47.1 22.4 75.0 10.4 8.7 50.3 27.0 77.5 13.4 6.5 49.2 35.8 79.8 12.0 9.1
NCR 52.2 36.9 82.0 15.3 12.4 52.9 31.3 81.6 17.2 9.2 46.1 10.4 80.0 5.4 2.2
CAR 41.7 8.3 78.5 5.7 3.5 44.6 9.6 77.5 6.3 3.3 45.9 14.5 78.4 11.0 4.3
ARMM 43.3 13.1 77.3 8.6 8.6 46.1 24.5 84.5 10.4 11.1 50.3 58.1 95.8 19.0 2.6
CARAGA 46.6 51.7 92.3 17.9 4.0 58.7 68.7 98.1 43.0 14.0 49.0 24.1 74.0 12.5 6.6
CALABARZON 47.3 29.7 80.6 13.8 10.2 46.4 23.7 72.6 11.8 9.1 45.6 11.7 70.2 7.2 2.3
MIMAROPA 48.3 19.1 71.8 10.5 4.6 54.1 41.9 86.4 15.2 16.8 50.9 36.8 80.2 9.3 7.6
PHILIPPINES 47.9 18.0 74.2 9.9 6.3 49.7 21.3 77.7 12.4 7.8 48.5 23.1 76.7 10.0 6.6
Notes: Authors’ calculations’ using data sourced from APIS, PSA
46
Annex Table A-3 Income Poverty and Multidimensional Poverty Profiles by Various Subpopulations: 2015
Subpopulation Total Population Percent
Contribution
Income Poverty
Headcount
Percent
Contribution
MPI
headcount
Percent
Contribution
MPI (in %) Percent
Contribution
a. Sector of
Employment of
Head
Agriculture
25,449,931 29.9 43.3 56.0 15.7 62.4 45.8 50.6
Industry
13,332,916 15.7 20.8 14.1 6.4 13.3 25.6 14.6
Services
31,633,201 37.1 12.4 19.9 3.3 16.1 15.7 22.5
Not employed
14,787,919 17.4 13.3 10.0 3.6 8.2 17.2 12.3
b. Region
Ilocos Region
5,136,000 5.1 13.1 3.1 3.6 2.5 20.3 4.2
Cagayan Valley
3,497,900 3.4 15.8 2.5 4.7 2.3 18.9 2.7
Central Luzon
11,098,900 10.9 11.2 5.7 2.1 3.2 10.8 4.3
Bicol Region
6,032,100 5.9 36.0 9.9 8.7 7.2 35.6 8.3
Western Visayas
7,704,399 7.6 22.4 7.9 9.8 10.4 38.5 11.5
Central Visayas
7,446,800 7.3 27.6 9.4 8.1 8.3 29.3 8.3
Eastern Visayas
4,537,200 4.5 38.7 8.0 9.2 5.7 31.2 5.8
Western Mindanao
3,759,323 3.7 33.9 5.8 12.4 6.4 39.7 5.4
Northern Mindana
4,706,700 4.6 36.6 7.8 7.7 5.0 27.3 4.9
Southern Mindana
4,963,100 4.9 22.0 5.0 8.0 5.5 31.4 5.7
47
Subpopulation Total Population Percent
Contribution
Income Poverty
Headcount
Percent
Contribution
MPI
headcount
Percent
Contribution
MPI (in %) Percent
Contribution
Central Mindanao
4,599,200 4.5 37.3 7.8 15.1 9.5 43.8 7.5
NCR
12,651,700 12.5 3.9 2.3 2.4 4.2 11.1 5.2
CAR
1,783,500 1.8 19.7 1.6 3.9 1.0 20.9 1.4
ARMM
3,706,900 3.7 53.7 9.1 29.3 14.9 75.9 10.4
CARAGA
2,716,700 2.7 39.1 4.8 7.3 2.7 28.1 2.9
CALABARZON
14,127,200 13.9 9.1 5.9 3.2 6.3 13.3 6.8
MIMAROPA
3,089,300 3.0 24.4 3.4 11.4 4.9 37.1 4.8
c. Location
Rural
57,982,846
57.1 29.8 78.9 9.7 77.4
33.0 74.3
Urban
43,574,077
42.9 10.6 21.1 3.8 22.6
16.3 25.7
PHILIPPINES 101,556,923 100.0 21.6 100.0 7.2 100.0 26.1 100.0
Notes: Authors’ calculations’ using data sourced from 2015 FIES, PSA